ISSN 0918~2365 CODEN NICKA3

Animal Science and Technology

Official Journal of Japanese Society of Zootechnical Science

President Seiki WATANABE

Vice president Teru ISHIBASHI Shigeru SUGANO

Editor-in-Chief H SHISHIDO

Editorial Board Y AKIBA T FUKLIKAWA T HINO C KANNO M KOMATSU M MORI

H TSUBONE

Executive Editors A OKANO S SAKAI

Membership includes subscription to the Journal and the privilege of submitling

reports for publication The annllal membership fee is 8000 yen

All correspondences concerning this Journal should be addressed to Japanese

Society of Zootechnical Science 201 Nagatani Corporas lkenohata 2-9-4 Taitoku

Tokyo 110 Japan

Abbreviation Anim Sci Techno (Jpn)

Formerly The Japanese Journal of Zootechnical Science

ut$ ffi =r yenR

fplusmn tIl A E3 $ ~ A ~ fT

A Ill iN xu iJt = iiI) A Ill filj fttj Jlf a=

~~jilll~ P rL BJl ~~jif)C ~ fiE k ~Jlf~tJr F~ IJ rJ if ~ J5J

fiii III Hi]-f[l ~ iJt ~~lJ$llGj Jlf ~ w telflj cplusmn

~~$mfJT T lID J~Hmij[[8rtBz~ilfTEl1LiIH3~ 7t~J - if 7 A 201 i5

maflJ El~~fG

Notice about photocopying

~i=lltlll ht~ff~Hlt] L tIJI~ ltlff~gthW In order to photocopy any work from this publication

gtjiI 0) fi11-lt- ~ -c sO-O)rIit1imiddot ~-lt-3tI -Co c l you or your organization must obtain permission from the

following organization which has been delegated for copyshy

7thi~~fnftl~ right for clearance by the copyright owner of this publicamiddot

T 101 iIgtIftlilIRII 9+41 1m) ] [7~Fl lion Phone 03-3415-4621 Fax 03-3403-1738 Except in the USA

The Copyright Council of the Academic Societies

7 fi I) bAfftfiJL$(t~nt1ji~t(t Fil[2-- 4[-pound Akasaka 9-chome Minalo ku Tokyo 1) Japan

Copyrigh Clearance Center Inc Phone BI-3~3475middot4621 Fax 81-3-3403~1738

222 Rosewood Dr ve In the USA Danvers -~A O~923 lSA Copyright Clearance Center Inc

Phone (508) 750 8400 Fa (5081 750-4744 222 Rosewood Drive

Danvers MA 01923 USA Phone (5081 7508400 Fax (508 750-4744

10 Jcj~ 67 ~10

-~~x (~X)

ili[t]7 It f I) A J i J Q i IJ~h~ ]ut t ~~L~tJ O)M-~t

XI -- Ilit-Xi 835

~EpoundRH)ilqII13 tTlJO)mEil Colpoda aspera IMT Q

tfljmt~~ ~tIMWi nx

iii77Hrr(tO)~ltl 0) 1ampit c nHt~JE -O)JJffl Agung PURNOMOADI bull j7sectmJB~b ~EfliEil

piill5i 851

l7-7-OJit1lg)amp~7 711lt r I)~ A GO)rM217 l )tJ

e~c$~ ~[ill1l 862

ii3f~ - (~X) 1JDIllIlt4gtr~tf1JDfrlsect L t~Eff J 3 O)Iamp-ftcgtri~tilii-J t l

~L~ Wi2 gil m illtlIfamp- 11 W nX~fji2 869

l 7middotmiddott jJ ~ 7 ~ ffl t t j~~jltrriJ QJlL40)(R~iJJEft0)1msectpound

0 bull ~f5D-1lPJ bull ltfliiHtiJr ~=~ 878

ry r I) gs i Jo it Q in ovo I) f( 7 i 7 ~ 3 ~ ffl t t r 7 A 7 L ~ 3

1ampIpoundT~J~llfrs9 c nJffifv0)J0f middot middot middot middot tt11ifA amp~5i ~ttiiIUli 882

-A5t~x

fIJ~Ipound-j ltm=GlDfiO)~tampitO)ltf( middot middot middotJt[lfiJ~z bull 11$i[K 886

~~51flT 0) l -yen IS It Q fjUJUtrIlJj~O) iE~IJampJampIiIamp IpoundT~--~ ~ampftffAiO)

Jfil middotmiddotmiddotlilgtJIJm~ e~ 7i jll1El3~ filjEl3($X bull LltlJiX (rp7rJgtc~ bull 1foJ~~ffll $~~plusmnyenj bull ~lJiXBJl~ lJiXrs9JEK JI~rEX 893

peR tIIJ Qisectt~ bull ~J~amp6O)~fl~)3IJ

~ bull [lfiJlJlrsectJ~ bull I tf_ middot 900

mrjJB-=f

sectllfJ~qJO) ry r I) JlfIJo It Q 1J1YJjpoundJt~ ffl t t[ 7 lid ~ 7 i 7 1ii1l(i-=fJEmO) l(~tIlmiddot middotmiddot middot middotmiddot middot middot middot middot middottt1lifA amp~lC ~HMUfi 906

~~gc$middotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middotmiddot middotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddot 910

Animal Science and Technology

Vol 67 No 10 October 1996

Contents

Regular Papers (in English with Japanese Abstract)

Oyama K Mukai F Determination of Optimum Mating Design With Conshystraints on Inbreeding Level via a Simple Genetic Algorithm 835

Kakiichi N Kitamikado A Sasamori T Tanaka Y Ishiwata Y Sakurai A Shimizu K Kamata S Toxicity of Several Insecticides Against Ciliate Colpoda asperamiddotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddot middotmiddotmiddotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middot middotmiddot middot middotmiddot middotmiddotmiddot middotmiddot 844

Purnomoadi A Kurihara 1 ishida T Shibata 11 Abe A Kameoka K Application of Near Infrared Reflectance Spectroscop) to Predict Fecal Composition and Its ese for Digestibility Estimation 851

Takahashi K Shirai K Hattori tvl Conjugated Film from Reconstruction of Collagen-Soluble Egg Shell lembrane Protein Matrix 862

Research Kotes (in English)

Fujihara T Maeda S Ogino T Matsui T Naruse H The Growth and Fat Nutrition in Lambs Fed a Diet with Treated (Spray-Dried) Beef-Tallow Supplement 869

Kuchida K Hamaya S Saito Y Suzuki M Miyoshi S Deveopment of a Body Dimension Measurement Method for Dairy Cattle by Computer Image Analysis with Video Camera 878

Muramatsu T Mizutani Y Okumura J Effects of Incubation Time and Injecshytion Sites on Gene Transfection Efficiency in Chicken Embryos by In Ovo Lipofection - 882

Regular Paper (in Japanese with English Abstract)

Hirooka H Matsumoto M A Comparison on Selection Approaches for Japashynese Brown Cattie Based on Profitability middot middot middot middot 886

Sunagawa K Takahashi H Iseki Y Higashi M Umeda S Yamashiro H Nakamasu F Akutsu H Hongo F Shinjio A Shiroma S Kawashima Y Effects of Excessive Vater Intake on Net Absorptions of Volatile Fatty Acids in Goats Exposed to Hot Environment 893

Fei S Okayama T Yamanoue M Nishikawa L Mannen H Tsuji S Species Identification of Meats and Meat Products by PCR 900

Rapid Communication

Muramatsu T Mizutani Y Okumura J Live Detection of the Firefly Lucifermiddot ase Gene Expression by Bioluminescene in Incubating Chicken Embryos 906

News and Announcements 910

Application of Near Infrared Reflectance

Spectroscopy to Predict Fecal

Composition and Its Use for

Digestibility Estimation

Agung PURtOMOADI Mitsunori KURIHARA

Takehiro NISHIDA Masaki SHIBATA I

Akira ABE and Ken-ichi KAMEOKA

FacuIty of Agriculture Tokyo University of Agriculture Setagaya-ku Tokyo 156

National Institute of Animal Industry Tsukuba Norin Kenkyu Danchi Ibaraki-ken 305

(Recei ved Decem ber 27 1996)

Abstract Two experiments were carried out to investigate the applicability of near infrared reflectance spectroscopy (NIRS) to predict the chemical composition of feces and thereby to estimate digestibility based on a lignin indicator method Ninety five fecal samples collected from digestion trials using dairy cattle were SUbjected to NIRS for the prediction of chemical composition Three methods to determine feed digestibility were compared namely by digestion trial (in vivo) by the lignin indicator method using data from chemical analysis (LiGLab) and by the lignin indicator method from NIRS prediction (LIGNIR) DigestIbility was evaluated for three groups of feeds based on the type of feedstuff used in the ration The groups were italian ryegrass only (lRO n 20) Italian ryegrass-concentrate ration (IRCT n = 16) and Italian ryegrass -steamed wood ration (IRSW n = 12) The rations were adjusted so as to meet the total digestible nutrients requirement of the Japanese Feeding Standard This study showed that fecal compomiddot sition could be accurately predicted by NIRS The values obtained by the NIRS prediction method for unknown samples and the respective values obtained by chemical analysis were highly correlated for acid detergent fiber crude fiber lignin and ether extract the correlation coefficients (r) were 098 098 097 and 096 respectively Correlation coefficients for crude protein organic matter and energy were 091 091 and 082 respectively With respect to digestibility estimation the value for the L1GLab and LIGNIR estimations and that for the in vivo were very similar The difference between the LIGLab and LIGNlR values and the in vivo value was below 3 the standard deviation of difference was less than 5 The results show that digestibility estimation using lignin determined by NIRS as an indicator was useful for the routine evaluation of nutritive values because it is simple fast and accurate

Anim Sci Technol (Jpn) 67 (10) 851-861 1996 Key words NIRS Fecal composition Lignin Digestibility Dairy cattle

A number of studies to determine the nutrishy out The success in predicting the chemical tive values of feed using near infrared refleshy composition of various feedstuffs by NIRS3

23

ctance spectroscopy (NIRS) have been carried advanced the application of this method to the

Present address Agriculture Forestry and Fisheries Research Council Ministry of Agriculture Forestry and Fisheries Chiyoda-ku Tokyo 100

Anim Sci Techno (Jpn) 67 (0) 851-861 851 1996

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Animal Science and Technology

Official Journal of Japanese Society of Zootechnical Science

President Seiki WATANABE

Vice president Teru ISHIBASHI Shigeru SUGANO

Editor-in-Chief H SHISHIDO

Editorial Board Y AKIBA T FUKLIKAWA T HINO C KANNO M KOMATSU M MORI

H TSUBONE

Executive Editors A OKANO S SAKAI

Membership includes subscription to the Journal and the privilege of submitling

reports for publication The annllal membership fee is 8000 yen

All correspondences concerning this Journal should be addressed to Japanese

Society of Zootechnical Science 201 Nagatani Corporas lkenohata 2-9-4 Taitoku

Tokyo 110 Japan

Abbreviation Anim Sci Techno (Jpn)

Formerly The Japanese Journal of Zootechnical Science

ut$ ffi =r yenR

fplusmn tIl A E3 $ ~ A ~ fT

A Ill iN xu iJt = iiI) A Ill filj fttj Jlf a=

~~jilll~ P rL BJl ~~jif)C ~ fiE k ~Jlf~tJr F~ IJ rJ if ~ J5J

fiii III Hi]-f[l ~ iJt ~~lJ$llGj Jlf ~ w telflj cplusmn

~~$mfJT T lID J~Hmij[[8rtBz~ilfTEl1LiIH3~ 7t~J - if 7 A 201 i5

maflJ El~~fG

Notice about photocopying

~i=lltlll ht~ff~Hlt] L tIJI~ ltlff~gthW In order to photocopy any work from this publication

gtjiI 0) fi11-lt- ~ -c sO-O)rIit1imiddot ~-lt-3tI -Co c l you or your organization must obtain permission from the

following organization which has been delegated for copyshy

7thi~~fnftl~ right for clearance by the copyright owner of this publicamiddot

T 101 iIgtIftlilIRII 9+41 1m) ] [7~Fl lion Phone 03-3415-4621 Fax 03-3403-1738 Except in the USA

The Copyright Council of the Academic Societies

7 fi I) bAfftfiJL$(t~nt1ji~t(t Fil[2-- 4[-pound Akasaka 9-chome Minalo ku Tokyo 1) Japan

Copyrigh Clearance Center Inc Phone BI-3~3475middot4621 Fax 81-3-3403~1738

222 Rosewood Dr ve In the USA Danvers -~A O~923 lSA Copyright Clearance Center Inc

Phone (508) 750 8400 Fa (5081 750-4744 222 Rosewood Drive

Danvers MA 01923 USA Phone (5081 7508400 Fax (508 750-4744

10 Jcj~ 67 ~10

-~~x (~X)

ili[t]7 It f I) A J i J Q i IJ~h~ ]ut t ~~L~tJ O)M-~t

XI -- Ilit-Xi 835

~EpoundRH)ilqII13 tTlJO)mEil Colpoda aspera IMT Q

tfljmt~~ ~tIMWi nx

iii77Hrr(tO)~ltl 0) 1ampit c nHt~JE -O)JJffl Agung PURNOMOADI bull j7sectmJB~b ~EfliEil

piill5i 851

l7-7-OJit1lg)amp~7 711lt r I)~ A GO)rM217 l )tJ

e~c$~ ~[ill1l 862

ii3f~ - (~X) 1JDIllIlt4gtr~tf1JDfrlsect L t~Eff J 3 O)Iamp-ftcgtri~tilii-J t l

~L~ Wi2 gil m illtlIfamp- 11 W nX~fji2 869

l 7middotmiddott jJ ~ 7 ~ ffl t t j~~jltrriJ QJlL40)(R~iJJEft0)1msectpound

0 bull ~f5D-1lPJ bull ltfliiHtiJr ~=~ 878

ry r I) gs i Jo it Q in ovo I) f( 7 i 7 ~ 3 ~ ffl t t r 7 A 7 L ~ 3

1ampIpoundT~J~llfrs9 c nJffifv0)J0f middot middot middot middot tt11ifA amp~5i ~ttiiIUli 882

-A5t~x

fIJ~Ipound-j ltm=GlDfiO)~tampitO)ltf( middot middot middotJt[lfiJ~z bull 11$i[K 886

~~51flT 0) l -yen IS It Q fjUJUtrIlJj~O) iE~IJampJampIiIamp IpoundT~--~ ~ampftffAiO)

Jfil middotmiddotmiddotlilgtJIJm~ e~ 7i jll1El3~ filjEl3($X bull LltlJiX (rp7rJgtc~ bull 1foJ~~ffll $~~plusmnyenj bull ~lJiXBJl~ lJiXrs9JEK JI~rEX 893

peR tIIJ Qisectt~ bull ~J~amp6O)~fl~)3IJ

~ bull [lfiJlJlrsectJ~ bull I tf_ middot 900

mrjJB-=f

sectllfJ~qJO) ry r I) JlfIJo It Q 1J1YJjpoundJt~ ffl t t[ 7 lid ~ 7 i 7 1ii1l(i-=fJEmO) l(~tIlmiddot middotmiddot middot middotmiddot middot middot middot middot middottt1lifA amp~lC ~HMUfi 906

~~gc$middotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middotmiddot middotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddot 910

Animal Science and Technology

Vol 67 No 10 October 1996

Contents

Regular Papers (in English with Japanese Abstract)

Oyama K Mukai F Determination of Optimum Mating Design With Conshystraints on Inbreeding Level via a Simple Genetic Algorithm 835

Kakiichi N Kitamikado A Sasamori T Tanaka Y Ishiwata Y Sakurai A Shimizu K Kamata S Toxicity of Several Insecticides Against Ciliate Colpoda asperamiddotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddot middotmiddotmiddotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middot middotmiddot middot middotmiddot middotmiddotmiddot middotmiddot 844

Purnomoadi A Kurihara 1 ishida T Shibata 11 Abe A Kameoka K Application of Near Infrared Reflectance Spectroscop) to Predict Fecal Composition and Its ese for Digestibility Estimation 851

Takahashi K Shirai K Hattori tvl Conjugated Film from Reconstruction of Collagen-Soluble Egg Shell lembrane Protein Matrix 862

Research Kotes (in English)

Fujihara T Maeda S Ogino T Matsui T Naruse H The Growth and Fat Nutrition in Lambs Fed a Diet with Treated (Spray-Dried) Beef-Tallow Supplement 869

Kuchida K Hamaya S Saito Y Suzuki M Miyoshi S Deveopment of a Body Dimension Measurement Method for Dairy Cattle by Computer Image Analysis with Video Camera 878

Muramatsu T Mizutani Y Okumura J Effects of Incubation Time and Injecshytion Sites on Gene Transfection Efficiency in Chicken Embryos by In Ovo Lipofection - 882

Regular Paper (in Japanese with English Abstract)

Hirooka H Matsumoto M A Comparison on Selection Approaches for Japashynese Brown Cattie Based on Profitability middot middot middot middot 886

Sunagawa K Takahashi H Iseki Y Higashi M Umeda S Yamashiro H Nakamasu F Akutsu H Hongo F Shinjio A Shiroma S Kawashima Y Effects of Excessive Vater Intake on Net Absorptions of Volatile Fatty Acids in Goats Exposed to Hot Environment 893

Fei S Okayama T Yamanoue M Nishikawa L Mannen H Tsuji S Species Identification of Meats and Meat Products by PCR 900

Rapid Communication

Muramatsu T Mizutani Y Okumura J Live Detection of the Firefly Lucifermiddot ase Gene Expression by Bioluminescene in Incubating Chicken Embryos 906

News and Announcements 910

Application of Near Infrared Reflectance

Spectroscopy to Predict Fecal

Composition and Its Use for

Digestibility Estimation

Agung PURtOMOADI Mitsunori KURIHARA

Takehiro NISHIDA Masaki SHIBATA I

Akira ABE and Ken-ichi KAMEOKA

FacuIty of Agriculture Tokyo University of Agriculture Setagaya-ku Tokyo 156

National Institute of Animal Industry Tsukuba Norin Kenkyu Danchi Ibaraki-ken 305

(Recei ved Decem ber 27 1996)

Abstract Two experiments were carried out to investigate the applicability of near infrared reflectance spectroscopy (NIRS) to predict the chemical composition of feces and thereby to estimate digestibility based on a lignin indicator method Ninety five fecal samples collected from digestion trials using dairy cattle were SUbjected to NIRS for the prediction of chemical composition Three methods to determine feed digestibility were compared namely by digestion trial (in vivo) by the lignin indicator method using data from chemical analysis (LiGLab) and by the lignin indicator method from NIRS prediction (LIGNIR) DigestIbility was evaluated for three groups of feeds based on the type of feedstuff used in the ration The groups were italian ryegrass only (lRO n 20) Italian ryegrass-concentrate ration (IRCT n = 16) and Italian ryegrass -steamed wood ration (IRSW n = 12) The rations were adjusted so as to meet the total digestible nutrients requirement of the Japanese Feeding Standard This study showed that fecal compomiddot sition could be accurately predicted by NIRS The values obtained by the NIRS prediction method for unknown samples and the respective values obtained by chemical analysis were highly correlated for acid detergent fiber crude fiber lignin and ether extract the correlation coefficients (r) were 098 098 097 and 096 respectively Correlation coefficients for crude protein organic matter and energy were 091 091 and 082 respectively With respect to digestibility estimation the value for the L1GLab and LIGNIR estimations and that for the in vivo were very similar The difference between the LIGLab and LIGNlR values and the in vivo value was below 3 the standard deviation of difference was less than 5 The results show that digestibility estimation using lignin determined by NIRS as an indicator was useful for the routine evaluation of nutritive values because it is simple fast and accurate

Anim Sci Technol (Jpn) 67 (10) 851-861 1996 Key words NIRS Fecal composition Lignin Digestibility Dairy cattle

A number of studies to determine the nutrishy out The success in predicting the chemical tive values of feed using near infrared refleshy composition of various feedstuffs by NIRS3

23

ctance spectroscopy (NIRS) have been carried advanced the application of this method to the

Present address Agriculture Forestry and Fisheries Research Council Ministry of Agriculture Forestry and Fisheries Chiyoda-ku Tokyo 100

Anim Sci Techno (Jpn) 67 (0) 851-861 851 1996

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

10 Jcj~ 67 ~10

-~~x (~X)

ili[t]7 It f I) A J i J Q i IJ~h~ ]ut t ~~L~tJ O)M-~t

XI -- Ilit-Xi 835

~EpoundRH)ilqII13 tTlJO)mEil Colpoda aspera IMT Q

tfljmt~~ ~tIMWi nx

iii77Hrr(tO)~ltl 0) 1ampit c nHt~JE -O)JJffl Agung PURNOMOADI bull j7sectmJB~b ~EfliEil

piill5i 851

l7-7-OJit1lg)amp~7 711lt r I)~ A GO)rM217 l )tJ

e~c$~ ~[ill1l 862

ii3f~ - (~X) 1JDIllIlt4gtr~tf1JDfrlsect L t~Eff J 3 O)Iamp-ftcgtri~tilii-J t l

~L~ Wi2 gil m illtlIfamp- 11 W nX~fji2 869

l 7middotmiddott jJ ~ 7 ~ ffl t t j~~jltrriJ QJlL40)(R~iJJEft0)1msectpound

0 bull ~f5D-1lPJ bull ltfliiHtiJr ~=~ 878

ry r I) gs i Jo it Q in ovo I) f( 7 i 7 ~ 3 ~ ffl t t r 7 A 7 L ~ 3

1ampIpoundT~J~llfrs9 c nJffifv0)J0f middot middot middot middot tt11ifA amp~5i ~ttiiIUli 882

-A5t~x

fIJ~Ipound-j ltm=GlDfiO)~tampitO)ltf( middot middot middotJt[lfiJ~z bull 11$i[K 886

~~51flT 0) l -yen IS It Q fjUJUtrIlJj~O) iE~IJampJampIiIamp IpoundT~--~ ~ampftffAiO)

Jfil middotmiddotmiddotlilgtJIJm~ e~ 7i jll1El3~ filjEl3($X bull LltlJiX (rp7rJgtc~ bull 1foJ~~ffll $~~plusmnyenj bull ~lJiXBJl~ lJiXrs9JEK JI~rEX 893

peR tIIJ Qisectt~ bull ~J~amp6O)~fl~)3IJ

~ bull [lfiJlJlrsectJ~ bull I tf_ middot 900

mrjJB-=f

sectllfJ~qJO) ry r I) JlfIJo It Q 1J1YJjpoundJt~ ffl t t[ 7 lid ~ 7 i 7 1ii1l(i-=fJEmO) l(~tIlmiddot middotmiddot middot middotmiddot middot middot middot middot middottt1lifA amp~lC ~HMUfi 906

~~gc$middotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middotmiddot middotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddot 910

Animal Science and Technology

Vol 67 No 10 October 1996

Contents

Regular Papers (in English with Japanese Abstract)

Oyama K Mukai F Determination of Optimum Mating Design With Conshystraints on Inbreeding Level via a Simple Genetic Algorithm 835

Kakiichi N Kitamikado A Sasamori T Tanaka Y Ishiwata Y Sakurai A Shimizu K Kamata S Toxicity of Several Insecticides Against Ciliate Colpoda asperamiddotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddot middotmiddotmiddotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middot middotmiddot middot middotmiddot middotmiddotmiddot middotmiddot 844

Purnomoadi A Kurihara 1 ishida T Shibata 11 Abe A Kameoka K Application of Near Infrared Reflectance Spectroscop) to Predict Fecal Composition and Its ese for Digestibility Estimation 851

Takahashi K Shirai K Hattori tvl Conjugated Film from Reconstruction of Collagen-Soluble Egg Shell lembrane Protein Matrix 862

Research Kotes (in English)

Fujihara T Maeda S Ogino T Matsui T Naruse H The Growth and Fat Nutrition in Lambs Fed a Diet with Treated (Spray-Dried) Beef-Tallow Supplement 869

Kuchida K Hamaya S Saito Y Suzuki M Miyoshi S Deveopment of a Body Dimension Measurement Method for Dairy Cattle by Computer Image Analysis with Video Camera 878

Muramatsu T Mizutani Y Okumura J Effects of Incubation Time and Injecshytion Sites on Gene Transfection Efficiency in Chicken Embryos by In Ovo Lipofection - 882

Regular Paper (in Japanese with English Abstract)

Hirooka H Matsumoto M A Comparison on Selection Approaches for Japashynese Brown Cattie Based on Profitability middot middot middot middot 886

Sunagawa K Takahashi H Iseki Y Higashi M Umeda S Yamashiro H Nakamasu F Akutsu H Hongo F Shinjio A Shiroma S Kawashima Y Effects of Excessive Vater Intake on Net Absorptions of Volatile Fatty Acids in Goats Exposed to Hot Environment 893

Fei S Okayama T Yamanoue M Nishikawa L Mannen H Tsuji S Species Identification of Meats and Meat Products by PCR 900

Rapid Communication

Muramatsu T Mizutani Y Okumura J Live Detection of the Firefly Lucifermiddot ase Gene Expression by Bioluminescene in Incubating Chicken Embryos 906

News and Announcements 910

Application of Near Infrared Reflectance

Spectroscopy to Predict Fecal

Composition and Its Use for

Digestibility Estimation

Agung PURtOMOADI Mitsunori KURIHARA

Takehiro NISHIDA Masaki SHIBATA I

Akira ABE and Ken-ichi KAMEOKA

FacuIty of Agriculture Tokyo University of Agriculture Setagaya-ku Tokyo 156

National Institute of Animal Industry Tsukuba Norin Kenkyu Danchi Ibaraki-ken 305

(Recei ved Decem ber 27 1996)

Abstract Two experiments were carried out to investigate the applicability of near infrared reflectance spectroscopy (NIRS) to predict the chemical composition of feces and thereby to estimate digestibility based on a lignin indicator method Ninety five fecal samples collected from digestion trials using dairy cattle were SUbjected to NIRS for the prediction of chemical composition Three methods to determine feed digestibility were compared namely by digestion trial (in vivo) by the lignin indicator method using data from chemical analysis (LiGLab) and by the lignin indicator method from NIRS prediction (LIGNIR) DigestIbility was evaluated for three groups of feeds based on the type of feedstuff used in the ration The groups were italian ryegrass only (lRO n 20) Italian ryegrass-concentrate ration (IRCT n = 16) and Italian ryegrass -steamed wood ration (IRSW n = 12) The rations were adjusted so as to meet the total digestible nutrients requirement of the Japanese Feeding Standard This study showed that fecal compomiddot sition could be accurately predicted by NIRS The values obtained by the NIRS prediction method for unknown samples and the respective values obtained by chemical analysis were highly correlated for acid detergent fiber crude fiber lignin and ether extract the correlation coefficients (r) were 098 098 097 and 096 respectively Correlation coefficients for crude protein organic matter and energy were 091 091 and 082 respectively With respect to digestibility estimation the value for the L1GLab and LIGNIR estimations and that for the in vivo were very similar The difference between the LIGLab and LIGNlR values and the in vivo value was below 3 the standard deviation of difference was less than 5 The results show that digestibility estimation using lignin determined by NIRS as an indicator was useful for the routine evaluation of nutritive values because it is simple fast and accurate

Anim Sci Technol (Jpn) 67 (10) 851-861 1996 Key words NIRS Fecal composition Lignin Digestibility Dairy cattle

A number of studies to determine the nutrishy out The success in predicting the chemical tive values of feed using near infrared refleshy composition of various feedstuffs by NIRS3

23

ctance spectroscopy (NIRS) have been carried advanced the application of this method to the

Present address Agriculture Forestry and Fisheries Research Council Ministry of Agriculture Forestry and Fisheries Chiyoda-ku Tokyo 100

Anim Sci Techno (Jpn) 67 (0) 851-861 851 1996

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Animal Science and Technology

Vol 67 No 10 October 1996

Contents

Regular Papers (in English with Japanese Abstract)

Oyama K Mukai F Determination of Optimum Mating Design With Conshystraints on Inbreeding Level via a Simple Genetic Algorithm 835

Kakiichi N Kitamikado A Sasamori T Tanaka Y Ishiwata Y Sakurai A Shimizu K Kamata S Toxicity of Several Insecticides Against Ciliate Colpoda asperamiddotmiddotmiddotmiddotmiddotmiddot middotmiddotmiddot middotmiddotmiddotmiddotmiddot middot middotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddot middot middotmiddot middot middotmiddot middotmiddotmiddot middotmiddot 844

Purnomoadi A Kurihara 1 ishida T Shibata 11 Abe A Kameoka K Application of Near Infrared Reflectance Spectroscop) to Predict Fecal Composition and Its ese for Digestibility Estimation 851

Takahashi K Shirai K Hattori tvl Conjugated Film from Reconstruction of Collagen-Soluble Egg Shell lembrane Protein Matrix 862

Research Kotes (in English)

Fujihara T Maeda S Ogino T Matsui T Naruse H The Growth and Fat Nutrition in Lambs Fed a Diet with Treated (Spray-Dried) Beef-Tallow Supplement 869

Kuchida K Hamaya S Saito Y Suzuki M Miyoshi S Deveopment of a Body Dimension Measurement Method for Dairy Cattle by Computer Image Analysis with Video Camera 878

Muramatsu T Mizutani Y Okumura J Effects of Incubation Time and Injecshytion Sites on Gene Transfection Efficiency in Chicken Embryos by In Ovo Lipofection - 882

Regular Paper (in Japanese with English Abstract)

Hirooka H Matsumoto M A Comparison on Selection Approaches for Japashynese Brown Cattie Based on Profitability middot middot middot middot 886

Sunagawa K Takahashi H Iseki Y Higashi M Umeda S Yamashiro H Nakamasu F Akutsu H Hongo F Shinjio A Shiroma S Kawashima Y Effects of Excessive Vater Intake on Net Absorptions of Volatile Fatty Acids in Goats Exposed to Hot Environment 893

Fei S Okayama T Yamanoue M Nishikawa L Mannen H Tsuji S Species Identification of Meats and Meat Products by PCR 900

Rapid Communication

Muramatsu T Mizutani Y Okumura J Live Detection of the Firefly Lucifermiddot ase Gene Expression by Bioluminescene in Incubating Chicken Embryos 906

News and Announcements 910

Application of Near Infrared Reflectance

Spectroscopy to Predict Fecal

Composition and Its Use for

Digestibility Estimation

Agung PURtOMOADI Mitsunori KURIHARA

Takehiro NISHIDA Masaki SHIBATA I

Akira ABE and Ken-ichi KAMEOKA

FacuIty of Agriculture Tokyo University of Agriculture Setagaya-ku Tokyo 156

National Institute of Animal Industry Tsukuba Norin Kenkyu Danchi Ibaraki-ken 305

(Recei ved Decem ber 27 1996)

Abstract Two experiments were carried out to investigate the applicability of near infrared reflectance spectroscopy (NIRS) to predict the chemical composition of feces and thereby to estimate digestibility based on a lignin indicator method Ninety five fecal samples collected from digestion trials using dairy cattle were SUbjected to NIRS for the prediction of chemical composition Three methods to determine feed digestibility were compared namely by digestion trial (in vivo) by the lignin indicator method using data from chemical analysis (LiGLab) and by the lignin indicator method from NIRS prediction (LIGNIR) DigestIbility was evaluated for three groups of feeds based on the type of feedstuff used in the ration The groups were italian ryegrass only (lRO n 20) Italian ryegrass-concentrate ration (IRCT n = 16) and Italian ryegrass -steamed wood ration (IRSW n = 12) The rations were adjusted so as to meet the total digestible nutrients requirement of the Japanese Feeding Standard This study showed that fecal compomiddot sition could be accurately predicted by NIRS The values obtained by the NIRS prediction method for unknown samples and the respective values obtained by chemical analysis were highly correlated for acid detergent fiber crude fiber lignin and ether extract the correlation coefficients (r) were 098 098 097 and 096 respectively Correlation coefficients for crude protein organic matter and energy were 091 091 and 082 respectively With respect to digestibility estimation the value for the L1GLab and LIGNIR estimations and that for the in vivo were very similar The difference between the LIGLab and LIGNlR values and the in vivo value was below 3 the standard deviation of difference was less than 5 The results show that digestibility estimation using lignin determined by NIRS as an indicator was useful for the routine evaluation of nutritive values because it is simple fast and accurate

Anim Sci Technol (Jpn) 67 (10) 851-861 1996 Key words NIRS Fecal composition Lignin Digestibility Dairy cattle

A number of studies to determine the nutrishy out The success in predicting the chemical tive values of feed using near infrared refleshy composition of various feedstuffs by NIRS3

23

ctance spectroscopy (NIRS) have been carried advanced the application of this method to the

Present address Agriculture Forestry and Fisheries Research Council Ministry of Agriculture Forestry and Fisheries Chiyoda-ku Tokyo 100

Anim Sci Techno (Jpn) 67 (0) 851-861 851 1996

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Application of Near Infrared Reflectance

Spectroscopy to Predict Fecal

Composition and Its Use for

Digestibility Estimation

Agung PURtOMOADI Mitsunori KURIHARA

Takehiro NISHIDA Masaki SHIBATA I

Akira ABE and Ken-ichi KAMEOKA

FacuIty of Agriculture Tokyo University of Agriculture Setagaya-ku Tokyo 156

National Institute of Animal Industry Tsukuba Norin Kenkyu Danchi Ibaraki-ken 305

(Recei ved Decem ber 27 1996)

Abstract Two experiments were carried out to investigate the applicability of near infrared reflectance spectroscopy (NIRS) to predict the chemical composition of feces and thereby to estimate digestibility based on a lignin indicator method Ninety five fecal samples collected from digestion trials using dairy cattle were SUbjected to NIRS for the prediction of chemical composition Three methods to determine feed digestibility were compared namely by digestion trial (in vivo) by the lignin indicator method using data from chemical analysis (LiGLab) and by the lignin indicator method from NIRS prediction (LIGNIR) DigestIbility was evaluated for three groups of feeds based on the type of feedstuff used in the ration The groups were italian ryegrass only (lRO n 20) Italian ryegrass-concentrate ration (IRCT n = 16) and Italian ryegrass -steamed wood ration (IRSW n = 12) The rations were adjusted so as to meet the total digestible nutrients requirement of the Japanese Feeding Standard This study showed that fecal compomiddot sition could be accurately predicted by NIRS The values obtained by the NIRS prediction method for unknown samples and the respective values obtained by chemical analysis were highly correlated for acid detergent fiber crude fiber lignin and ether extract the correlation coefficients (r) were 098 098 097 and 096 respectively Correlation coefficients for crude protein organic matter and energy were 091 091 and 082 respectively With respect to digestibility estimation the value for the L1GLab and LIGNIR estimations and that for the in vivo were very similar The difference between the LIGLab and LIGNlR values and the in vivo value was below 3 the standard deviation of difference was less than 5 The results show that digestibility estimation using lignin determined by NIRS as an indicator was useful for the routine evaluation of nutritive values because it is simple fast and accurate

Anim Sci Technol (Jpn) 67 (10) 851-861 1996 Key words NIRS Fecal composition Lignin Digestibility Dairy cattle

A number of studies to determine the nutrishy out The success in predicting the chemical tive values of feed using near infrared refleshy composition of various feedstuffs by NIRS3

23

ctance spectroscopy (NIRS) have been carried advanced the application of this method to the

Present address Agriculture Forestry and Fisheries Research Council Ministry of Agriculture Forestry and Fisheries Chiyoda-ku Tokyo 100

Anim Sci Techno (Jpn) 67 (0) 851-861 851 1996

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Pl]010lJl 1 L RII LH ISIDA SlllBA T 1 ABE and 1IEOI

prediction of nutritive values such as lolal dimiddot

gestible nutrients (10) and avaUable cnerg

conleltsmiddot11 these stlldes wcre carried

out directly ith feeds in which the in li1O

digestibilit y as Y Known Howcmiddotmiddot

cr digcstibilitmiddot is affected I) the condition of

the animal used variation bel(cl1 animals as

ell as le(I Therefore it is

IJIc for the same ration fed to different animah

to haH differcnt aild

gestibility should be

The ideal way to measure is by

the digestion trial 111ethod in which feed intake

Zll1d the nutrients ingcst((j can Ile accurately

measured However this mcthod is

and laborious n alternatic method in dimiddot

gcstibitmiddot measurelllents is the indicator

method in which I is used as an indicator

This ll1ethod docs lot the total collecmiddot

tion of feces but only needs a representative

fecal sample Oilly by llleasuring the percentmiddot

age of lignin ill feed and feces can the

bility of the feed be estimated

Gien that the chemical composition of feeds

are eaSily and routinely using ]IRS

it may be worth to usc the same apshy

proach to estimate the chemical composition of

feces One extra ad vantage of this approach is

that it takes into account the fraction of inmiddot

gested nutrients and energy lost in the feces

which accounts for the part of feed

losses_ In ruminants these losses reach 40shy

50 and 20-30 in case of roughage and conmiddot

centrate Combined with the

efficient indicator method NIRS may be used

to generate a faster and more accurate estimashy

tion of feed digestibility from individual

animals

The aitrls of this study were to validate the

usc of IRS for the prediction of the chemical

of feces and to c-aluate the accushy

racy of estimation using lignin

by IRS

Materials and Methods

Experiment l

fecal samples from a digestion

trial dairy cattle were used in this study

All cattle were fed at the total digestible nutrishy

ents (TD]) requirement leel of the Japanese

Standard for Dairy Cattle The

samples were dried at tiOT for 48 hours before

were ground ith a Wiley mill to pass

through a 100-mm screen These processed

samples were used for chemical analysis to

determine organic matter (OM) crude protein

(CP) ether extract (EE) crude fiber (CF)h acid

detergent fiber (ADFI acid detergent lignin

(lignin) and silica n as well as energy

Analysis by NIRS was done by a Pacific Scimiddot

entific (Neotec) model 6500 (Perstorp Analytmiddot

ical Silver Spring 10) instrument equipped

with lSI software (lnfraSoft International Port

Matilda PAl for analysis_ The samples were

scanned over the range 1100-2500 nm The

spectral data were collected at wavelength inshy

tervals of 2 nm and the 700 data points obmiddot

tained for each sample were stored in the comshy

puter as absorbance values_ These values

were expressed as log 10 (lR) where R is the

reflectance The second derivative of log lR

values were used to derive relationships with

the chemical compositions 23 To achieve the

optimum wavelengths for predicting the comshy

ponents a stepwise multiple-linear-regression

program was used

Forty-nine samples chosen randomly from

the 95 fecal samples above were used as a

standard sample to develop a NIRS calibration

equation This equation was then used to preshy

dict the chemical contents of the 46

fecal samples (unknown The camiddot

libration was done with a maximum of four

The least number of waveshy

lengths was used for measurement when an

increase in the number of wavelengths gave no

increase in of the R value nor any

decrease in the standard error of calibration

85L

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

DigltSlilJilily estimation b IRS predicted data

Experiment 2

ft is well known that prediction b iIRS will

be maximal if the calibration set samples and

unknown samples are the same or of high homogeneity This is because the feedmiddot

stuffs were separated in to two big groups to

develop the calibration equation

Twenty-five samples of forage including

hay silage and steamed wood and 10 samples

of concentrate including grains and soybean

were used to develop an lIRS calibration equamiddot

tion for forage and concentrate respectimiddotely

Due to the limited number of feed samples

used the validity of the developed equations

was not tested Both equations were directly

used to determine Ovl cp EE CF ADF lignin

and energy contents for 10 feedstuffs (7 Italian

ryegrass 2 concentrate and 1 steamed wood

samples) These feedstuffs were the main

components used to make the 48 rations with

known in vivo digestibilities Subsequently

the values observed from the 10 feedstuffs

were used to calculate the chemical composimiddot

tion of the 48 rations Baseo on types of

feedstuffs used in the rations three groups

were designed for evaluation Group I was

Italian ryegrass only (IRO) composed of three

kinds of Italian ryegrass fed to dairy cattle (n

20) Group 11 was a mixture of Italian ryegrass

and concentrate (IRCT) composed of () a commiddot

bination of Italian ryegrass and soybean (2) a

com bination of Italian ryegrass and commershy

cial formula feed and (3) a combination of Italshy

ian ryegrass and soybean and commercial forshy

mula feed These rations were fed to dairy

cattle (n 16) Group III was a mixture of Italshy

ian ryegrass and steamed wood (IRSW) at 5

and 55 fed to 12 dairy cattle The separation

of rations for evaluation was made to reduce

the influence of feedstuff type on digestibility

estimation because of different lignin

recoveries These rations were allowed at

the TDi requirement level of the Japanese

Feeding Standard for Dairy Cattle20Jbull Finally

the digestibility of each ration was calculated

by the lignin indicator method using the

formula

digesti bili t Y 100 - lOO x ( lignin in

feeds lignin in feces)

x( nutrient in feces

nutrient in

Two approaches were applied to estimate

digestibility by the lignin indicator method

and both were compared with in lilo digestibilshy

ity values for their accuracy_ The first apshy

proach was to calculate the chemical composmiddot

tion of feed and fecal samples obtained from

chemical analysis (LiGLab) The second one

was to calculate those samples from iIRS data (LIGNIR)

Digestibility values estimated using LIGLab

LIGNIR and in vivo were compared for ()yl CP

EE CF ADF and energy The differences bemiddot

tween individual digestibility values estimated

by LIGLab or LIGNIR and in vivo were calculatshy

ed These differences were tested for signifimiddot

cance using paired t-tests28 )

Results and Discussion

Experiment 1 The ranges and means of all constituents for

the standard and unknown samples are shown

in Table I Regarding the range of composimiddot

tion values of the standard samples were

slightly wider than those of the unknown

samples

Wavelengths observed for calibra hon of

standard samples are listed in Table 2 With

respect to organic matter the 1st to 4th absorbshy

ance wavelengths were found at 2418 2206

1144 and 1956nm respectively The first abshy

sorbance was in the 2410-2460 nm region conshy

sidered to be the second overtone of C-H deformation9) and this was related to cellulose

Another report) found that the first two waveshy

lengths were absorbance values for the detershy

gent fiber component The 3rd wavelength

was close to 1I43nm characteristic of aromatshy

ic structures25 which is thought to be

The 4th wavelength which was close to 19f30

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

PURNOtvlOADI KURIHARA KISHIDA SHIBA TA ABE and KAMEOKA

Table 1 Means and ranges of constituents for calibration of standard samples and prediction of unknown fecal samples

Standard (n=49) Unknown (n 46) Constituents

Mean Range Mean Range ---~ --------~ ---_ _-__--__shy

Organic matter 851 662-94 85 7 708945

Crude protein 126 4 3 120 57-238

Ether extracts 1 9 05-30 18 04-35

Crude fiber 299 2-~59 31 187-41 2

Acid detergent fiber 475 4-68 1 08 34 4-64 2 Energy 4 42 276-4897 4439 3756-5065

Lignin lAo 96-253 145 9 6

Silica 6 3 0917 4 59 0 4 --__-----__- _

n number of samples

Table 2 Wavelengths used and correlation coefficients of each consituent for fecal calibration 0 standard and prediction of unknown fecal samples

Wavelength (nm) Standard (n=49) Unknown (n = 46) Constituerlts

1st 2nd 3rd 4th R SeC SeP

Organic matter 2418 2 205 956 O 957 29 09Il 0 35 Crude protein 2 2090 2 396 0988 070 0912 009 Ether extracts 1 470 2 ~ 0 983 014 0956 002 Crude tiber 332 1 2 1 0983 22 0976 O 19 Acid detergent fiber 300 330 O 1 45 0979 O 12 Energy 2420 1 572 C35 1 602 0 7Q6 e 2L Lignin 2 388 1 76e 2 318 0 971 07 0972 O

Silica 2 1372 1 54 0925 86 0792 006

0) R correlation coefficient from multiple regression r correlation coefficient from simple regression SeC standard error of calibration SeF standard error of prediction

nm was for protein25l bull

Three wavelengths at 2186 2090 and 2396

nm were recorded for the calibration of CPo The 1st wavelength was protein at 2180 nm25

)

The 2nd wavelength was considered to be celshylulose at 2088 nm25) while the 3rd wavelength

at 2396 nm was close to the 2380 nm of hemicell ulose25 )

Three wavelengths at 1470 2334 and 1690

nm were recorded for the calibration of EE

The 1st wavelength was 19nm different from

the 1489 nm of cellulose25) The 2nd and 3rd

wavelengths were very close to 2336 and 1685 nm which are known to be the absorbance of cellulose and lignin2225) respectively These

wavelengths demonstrated that dominant

fecal fibrous components were used in

determining ether extract Ether extracts in feces is the smallest constituent and does not

clearly appear in log IfR Consequently the

wavelength properties of EE might be conshy

cealed by the more dominant spectrum of celshy

lulose Crude fiber (CF) was well predicted by using

four wavelengths at 2332 18202452 and 1954

nm The first two wavelengths were close to

the absorbance of cellulose at 2336 and 1820 nm 25

) The 3rd and 4th wavelengths were

close to the 2461 nm absorbance of starch and 1960 nm absorbance of protein 2S

) The first

854

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Digestibility estimation by NIRS predictcc data

t vo wavelengths are known to be the most

important as cellulose is t hc most important

component of CF The 3rd wavelength that

of starch was considered to be important as

starch is structurally similar to cellulose The

4th wavelength that of protein apparently has

no relation with CF and was simply used to

improve the correlation coefficient in this reshygression

Only two wavelengths at 2300 and 1330 nm

were used for the calibration of ADF The

first wavelength was close to the 2294 nm of neutral detergent fiber (fDF)23) The second

wavelength was related to that of hemishycellulose at 1360 nm) Hemicellulose vith

cellulose and lignin form NDF but hemishy

cellulose does not form a part of ADF which

contains just cellulose and lignin The apshy

pearance of KDF or hemicellulose absorbance

in ADF was related to the fact that both cellushy

lose and hemicell ulose are carbohydrates in

which they have the same absorbance region

For lignin three wavelengths were observed which were 2388 1760 and 2318 nm The 1st

and 2nd wavelengths were close to the

reported absorbances of in vitro dry matter digestibility (lVDMD) at 2386nml) and at 1759

nmm respectively The 3rd wavelength was

considered to be hemicellulose which is usualshyly observed at 2314 nm 25 )

The four wavelengths observed for energy were 2420 1572 2036 and 1602 nm The first

wavelength was the same as the first waveshy

length of organic matter which was considshy

ered to be cellulose The 2nd wavelength was

close to either the reported 1570 nm of cell waIl6) or the 1580nm of cellulose25) The 3rd

wavelength at 2036 nm was 14 nm different

from 2050 nm which is the absorbance of protein 3J The fourth wavelength at 1602nm

was considered to be the absorbance of organic

matter That was close to the 1606nm absorbshyance of organic matterS) and the 1610 nm abshy

sorbance for IVDMD the first wavelength from five wavelengths reported by Holechek et arIS)

The four wavelengths observed here were in

agreement with the fact that organic matter is closely related to energyS

For silica three wavelengths were observed

at 2220 1372 and 1694 nm The first waveshy

length was close to the reported 2212 and 2214 nm of the digestible fraction of cell waIl3

)

The 2nd and the 3rd wa velengths were close to

the 1365 nm of ceilulose25J and 1685nm of Iignin25) respectively

Correlation coefficients (H) between the

values obtained by chemical analysis method

and those obtained by NIHS for all constituents

in the standard samples were observed to be

more than 090 The highest R 0988 was found for CP middot-hile the lowest 0925 was for

silica As presented in Table 2 H values for

the fibrous fractions were very close to one

being 0983 0987 and 0971 for CF ADF and

lignin respectively High H values were also

found for OM EE and energy contents being

0957 0983 and 0931 respectively For validishyty of equations developed the prediction equashy

tions in the standard samples were tested for

further application

The prediction equation obtained from the

standard samples was used to determine the

constituents of 46 unknown samples Except

for CF and lignin the simple correlation coeffishyden t (r) and standard errors (SeP) were genershy

ally lower than those for the standard samples_

Because of the higher similarity of R values for

lignin in the standard and unknown samples it

may lead to a possibility of lignin as an indicashy

tor for digestibility estimation The decrease in the correlation coefficient

values of ADF EE OM and CP observed from

the calibration of standard samples compared to that from the prediction of the unknown

samples was small A remarkable decrease of

the correlation coefficient was observed for

silica (from 0925 to 0792) and energy (from

0931 to 0819) This result shows that silica

was poorly predicted It agrees with the

finding that minerals are known to have no

855

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

PURNOMOAD KCRJRA ISIIIDA SHIBATA ABE and KAlEOKA

absorption bands in the ncar infrared region) 3 In forage high R values were found for Oc1

In case of energy using only four wavelengths CP CF ADF and lignin ranging from 0978 to

which is the maximum capacity of instrument 0986 Lower R values were found for EE and

used may not satisfy the requirements for energy being 0834 and 0770 respectively In

fecal prediction because of spectral complicashy addition in concentrate high R values were

tions resulting [rom the wide possibility of also found for OMCP EE ADF energy and

combinations This was due to the fact that lignin ranging from 0992 to 0997 A lower R

energy is the heat contributed from all chemishy value 0907 was found only [or crude fiber

cal components in the samples The high calibration R values for lignin in both

Reviewing the absorbances observed the forage and concentrate suggests that lignin in

wavelengths appropriate for feces were someshy feed may be well predicted and can be used for

what different from that usualy observed for digestibility estimation based on lignin

feed The wavelengths obsened for all conshy The chemical compositions of 48 rations

stituents in feces were dominantly in the cellushy classified into three groups IRO lRCT IRS

lose spectra for predicting chemical composimiddot are presented in Table 4 The chemical comshy

tion Reference to wavelengths obtained only position of the feedstuff groups were calculatshy

from feed made it difficult to conclude hether ed using the data obtained from chemical anal-

the wavelengths obtained were appropriate for and from ~IRS prediction of the 7 roughshy

feces However the obtained wavelengths ages 2 concentrates and 1 steamed wood comshy

were still related to the chemical composition ponents Referring to mean values only small predicted differences between the two methods of calcushy

Experiment 2 lation were observed for all constituents Relshy

The ranges and means of feedstuffs for NIRS atively large differences were observed only in

calibration in this study are presented in Table ADF of IRO and CF of IRSW being 17 and

Table 3 v1eans ranges and wavelengths used for )iIRS calibration of feedstuff samples)

Vavelength (nm) Cons tit uents Mean Range R SeC

1st 2nd 3rd 4th

Roughage (n 25)

Organic matter J 9-99 2 0978 51 Crude protein Q~ L C 4 2 l88 228 0980 054 Ether extracts 26 C8~3 4 584 285 220 040 0834 046 Crude fiber 31 247-478 2 18 1505 2360 0984 01

Acid detergen t 11 ber 39 308-581 740 2378 O 1 75

Energy 4 305 4 177-4770 1 1 692 2 ~ 964 O 770 83 Lignin 62 0-15 2378 670 O 981 064

Concentrate (n= 10) o~Organic rna t ler 93 L 3 L 258 490 2032 0922 007 ~tCrude protein ~ 9 7 2132 0993 1

Ether extracts 7-35 1974 2 2446 O 009

Crude fiber t t 1 1 1 9D7 o )4 1)

Acid detergen t fi ber 7 8-136 1 212 2 O 995 022 Energy ~ 0997 121 Ligrin 1 8 _810 O 992 O

R correlation coefficien~ from muliple regression SeC standard error of calibration

856

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Dige~tibiit cslimction by 0J1RS predicted data

Table 4 lcans of chemical composition of the ralion groups used in this study calculated with

chemical analysis data and lIRS predicted data (DImiddot

IRO n ~O) IRCT (n~ 16) IRSW (n =12) Constltucnts

Lab -IR Lab KlR Lab

g)

Crude protein 9

Organic maller o 9

Ethcr extracts 5

Crudc fiber 8 C

Acid detergent fiber 3 43 I

Encrgy 4 r4 1() 4115

Lignin 5 ~i 5

Feed composition calculation 0 48 rations used tht data of 7 roughage 2 concentrate and steamed wood eed obtained rrom chemical anclysis (Lab) and IRS pediclion (JEI IRO Ilalian rycgrass on]

IRCT llaLn ryegrass-co1centrales IRS llalian ryegrass steamed wood 1 Energmiddot ex pressed in calg D1

12 respectively Other constituents were

below 06 For energy a large difference

was only obsened for IRS being 55 calg

The mean digestibility calculated from the

three methods as well as the standard deviashy

tion of difference (SDd) between both LlGLab

o[ LIGNIR and that in vivo are shown in Table

5 The ealuation was done for 3 groups

separated on the basis of type of feedstuffs

In general the estimated values were slightshy

ly higher than the in vivo values A comparishy

son between LIGLab and LIGlJR for the IRO

and IRSW groups indicates that the LlGLab

estimated value was higher than that of

LlGlIR However the contrary was observed

for the IRCT group These figures may be

caused by the NIRS predicted data for lignin of

the rations (Table 4) Due to the equation for

digestibility estimation used the estimation

value will be lower if the value of lignin of

feces is low or lignin of feeds is high

For a comparison of the means between the digestibility estim2ted by LIGlIR and in vivo

the difference observed in OM and 01 were 23

lA 20 and 30 11 17 for the IRO IRCT

IRSW groups respectiely These differences

are relatively small For CP and EE digestibil shy

ity the differences of means were 12 16 58

and 09 05 48 [or the IRO IRCT ald IRSW

groups respectively The different values obshy

served for the IRO and IRCT groups were

small but were relatively high for the lRSVi

group Small differences were observed in CF and ADF digestibility being 05 06 28 and 21

03 12 for tIte IRO IRCT and IRSW groups

respectimiddotey For the digestibility of energy

the differences of means for IRO IRCT and

IRSW groups were 2A 15 and 16 respectiveshy

ly From these differences the digestibility

estimated with LIGNIR were very similar to

the values obtained in vivo excrpt for the dishy

gestibility of CP and EE for the IRSW group

For comparison Penning and Johnson 2S using

acid insoluble ash (AlA) as a marker observed

the mean differences of OM digestibility at 09

and 18 for sheep fed ryegrass at 15 and 25 g

per kg live weight respectively

Referring to the standard deviation of differshy

ence (SDd) between the estimation val ues of

LlGLab or LIGlIR and that in vivo the values

were generally below 5 for the IRO and IRCT

groups except for the digestibility of CP of the

IRO group For the IRSW group the SOd was

relatively high The accuracy of digestibility

estimations of similar studies to the present

experiment were expressed by the residual

standard deviation (RSD) which was observed

as the differences between the in vivo value (nd

857

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

------

PUH0110ADI KllRIILR lISlllDA SHIBATA ABE and KAMEOK

Table 5 Digestibility value estimated from LIGLab and LIGNIR in comparison with the in ito method

LIGLab LIGNlR

Mean SOd Mean SDd Mean Range

Dry matter

Organic matter

Crude protein

Ether extracts

Crude fi ber

Acid detergent fiber

Energy

Dry maller

Organic matter

Crude protein

Ether extracts

Crude fIber

Acid detergen 1 f1 ber

Energy

Dry -latter

Organic lalter

Crude poteh

Ether extracts

Crude fiber

Acid detergent fiber

Energy

Italian 1

4

4 4 lt1

4

~

rycgrass only (lRO

6middot 426~

5 4middot 4345

I~~ 6 118

55 2 3243

48 5 4 505

9 4827 88 42~7

n=20)

51 9 39 1~68

544 413-717 443 32

553 542-74

480 31 2-68 3

390 227-599

i1 2 381-688

----- ------~ Italian ryegrassconcentra tes (IRCT n ~ 16)

J~2 735 2255 72 ]

7S Smiddot 2 110 74 7

3 102 n 1

2 2 5~5 77 7

5l 9

555 47 8 ~ CO2 _lb 1 ~

----~~~ Ilalian ryegrasssteamed wood (IRS n=12)

3CG 569 ~ 586

r- ] 37 9 L 340 663

7 8 246 579 9middot 0 488

60 5 5 748 549

553-77 6

589-790 572-831

53 2-82 1

35 7~56 2 658~77 2

~-----

445-674 451-702 159-550 62 2-7L 8

340-71 8 299-624 401-668

~--~

) Significant (Plt 005) ) significant (PltOOl) SOd standard deviation of difference between estimation

values of the L1GLab or L1GiIR and the in vivo Dry feces)

the estimation adjusted value resulting from

the regression equation These RSD values

represented the estimation variation

associated with the techniques Standard demiddot

viation of differences and RSD are comparable

because they are both calculated by the differshy

ence between aiues determined in vivo and

those estimated The RSD of dry matter (D~1)

digestibility estimated using lignin determined by three methods ranged from 24 to 45 1

7)

while the same estimation made using cellulase

digestion methods was in the range of 25 to 27 111

bull

For Ol1 digestibility avaratne et alI

858

matter digestibility = 100 ~(lOO lignin feedlignin

the prediction of roughages (grass

and straw) by various methods These

methods used AlA as a marker (RSD=406)

rumen fluid (RSD 378) pepsin-cellulase (RSD

509) and nylon bag methods (RSD=408) If

the is done based on type of feed

RSD values observed for the grass group using

AlA marker rumen fluid pepsin-cellulase and

nylon bag methods however were lower by

178 166 140 and 124 than the results of the

present experiment respectively The estimamiddot

tion of OM digestibility in this was also

similar to that reported by Aerts et alP who

used two-stage in vitro method

0

Dgestibiliv estimation b NIHS predicted data

A relatively high SDd was obsened for dishy

gestibility of CP in the IRO and IRS groups

and for digestibility of energy in the IRSW

group These estimations are less accurate

than other components although this is

statistically not significant The high SDd of

CP digestibility observed in the IRO and IRSW

groups were merely caused by the wide range

of CP digestibility for the low CP content in the

feed (both groups were below 10 CPl Homiddot

ever eval ua tion of CP can be neglected for

ruminants due to the involvement of non proshy

tein nitrogen (NPN) in the calculation

Meanwhile the high SDd for digestibility of

energy in the IRSW group was considered to

be influenced by the poor energy prediction of

the feedstuff in that group (see Table 4) A

comparison of observed SDd with the RSD

reported from the various methods above indishy

cates that the estimation method used in this

study was apparently faorable

The most favorable estimation of digestibilishy

ty in the IRCT group as indicated by the smallshy

est bias and SDd among the rations was relatshy

ed to the accuracy of lignin prediction of

feedstuffs Since usual farming management

would use a com bination of roughage and conshy

centrate this digestibility estimation shows

the applicability for farms To improve the

accuracy of this method a larger sample

number and wider range of feedstuffs for

developing NIRS prediction equations should

be considered

In practice only DM digestibilit y3c or mvl

digestibilit y2) would be used for ealuating

feeding management The differences beshy

tween the lignin indicator methods and the in vivo method were related to the fact that lignin

was partiall y digestible lG 1I an Soest3G )

noted that the limitation of lignin as a

digestibility indicator is limited because of the

large inter-species varation However this

can be overcome by llsing fairly similar

feedstuffs although under these condiLons

standard error is still about 3 of digestibility

Thus the results above show that this

bility estimation method is sufficiently reliable

Digestibility estimation by LIGNIR shows

the potential for individual and routine measshy

urement Near infrared reflectance spectrosmiddot

copy prediction of the chemical composition of

feces and feed as well as the application of

iIRS prediction for digestibility estimation

mav be useful for the evaluation of feedstuffs

on a practical farm basis provided some corshy

rection factors or eq uations are devised to minshy

imize the difference between in vivo and that

estimated by LIGiIR

Acknowledgements

The authors wish to thallk Dr F Terada for

reading the manuscript and for his valuable

comments Dr -1 Amari for assisting in the

and 1s 1 Someya for assisting

with the chemical analysis The authors are

also indebted to Dr Muladno for assisting with

the writing

References

1) Abe A Feed analyses based on the carbohyshydrates and its applisation to the nutritive value of feeds Mem lat Inst Anim Ind No 2 23-29 National Institute of Animal Industry IIAFF TSJkuba Japan 1988 Cn Japanese)

2) ertsJ De Brabander DL Cotln BG Busse FX Comparison of laboratory methods for predcting the organic matter digestibility of forages Anim Feed Sci TechnoL 2 337349

1977 3) Amari 111 Abe A Tano R 11asaki S Serizawa

S Koga T Prediction of chemical composition and nutritive value of forages by near infrared reflectance spectroscopy L Prediction of chemical composition J Japan Grassl Sci 33 219-226 (in Japanese with English summary) 1987

4) American Society of Animal Science Tech~ niques and Procedures in Animal Production Research American Society of Animal Scimiddot ence 179 Albany 10 NY 1963

5) Armstrong DG Evaluation of artificially dried grass as a source of energy for sheep 11 The energy value of cocksfoot timothy and two

859

ITRtOOcDL K[1RIILH ISIIIDA SIlIl3ATA cBE and KAt1EOKA

strains of rye grass at middotaryin stages 0 r11C1tu 18) Jvlorimoto II Dobutsu eiyo shikenho (Experi

rity J Agric Sci Camb 62 399 1961 mental method of animal nutrition 280

6) Barnes RJ ear infra rcd spectra of ammona 119 Yokendo Tokyo 197L (in )apancs()

treated straw and isolated cell walls nim 191 -lurray L Williams Pc Chemical principles of

Feed Technol 21 209211)1988 near infrared technology in near infrared

7) Bondy AA nimal nutrition 3)~ icy technology in agricultural and food industries

Intersciencc PublicaticJr) John Wiley 6 Sons Williams Pc Norris KH 3031 American asso

Ltd Chichester lew York BrisiJane Toronto ciation of cereai chemists Inc SL Paul l1inne

Singapore 1987 tS1990

8) Coelho 1 lembry FG Barton FE Saxton 1 20) National Research Council of Agriculture For

A cornparison of rtdcrobia enrymltlcic chemi estry and Fisheries Japanese Feeding Stand

cal and nearinfrared reflectance spectroscopy ani fer Dairy Catlie 66-68 NRCAFF Ministry

methods in forage ealuation nm Feed Sci of Agriculture Forestry anc Fisheries 1987

TechtlD 20 2192]1 1988 21) aaratne IlYRG Ibrahim lll Schiere JB

9) De Boccr JL Cocn BG anacker JM Comparison of four techniques for predicting

Boucque Ch The use of nirs t) prwlic the digestibilit of ropical feeds Anim Feed

chemical composition and the energy liue Techno 29 bull 20921 1990

compound feeds for catllc nim Feed Sci 22) [on sOllchi koukai Soushiryou no

Techno 51 243-251 1995 hinshilsll hyouka gaid()~)Ukkujikuli shiryou

10) Elam CJ Dinis RE Lignin excretion [l cCluie hinshitsll houka kcnkyuukiiher~ 1

fed a mixed ~atjon J Anim Sci ~Q 4il(j KyoUa Toko 1995 In Japanese)

1961 0orris KII Barnes RF loorc JE Shen S

11) El RE Kane E Jacobson ( loore LA Predicting forage quaE by infrared

Studico the compOoilic)l of Ignin i~oiaed ctance spectroscopy 1 Anim Sci 43 887-889 from orchard grass hay Cll at four stages of 1976

maturit and frum the corre~polldlllg cces ) 211 Orskm ER Protein nutrition in ruminants 2 Dai Sci 36 3middot6-lJi nel edition 89 157 Academic Press Ltd

12) Givens D1 Baker C dalllson I CjSS R Londo] 1992

Inlluence of gn)wltl lype and on the 25) Osborne EG Fearn T Ncar infrared spectros

prediction of the Illcabolisablc conent copy in food analysis 3740 133 Longman

of herbage by nearinflued rellectance spec Scientific amp TechnicaL Longman Group CK

troscopy nim Feed Sci Techno 37 281 Limited Ergland 1986

295 1992 26) Penl1ing PD Johnson RIL Th of internal

13) lolechek JL Shenk JS YaH2 1 Arthun [) markers to estirna~e herbage digestibility and

Prediction of forage qualit using ncar infra intake I Potentiaily indigestible cellulose and

red rellectanc( spectroscopy on (sophageal insoluble ash J cgric Sci Camb 100

fistula samples from cattle or mountain r2nge 127Ll1 1983 J Anim Sci 55 971 97S 1982 Shenk JS Westerhaus 0 Hoover MR Anal y

14) Jones DIH Hayward IV A cellulase digestion sis of by infrared relleclance J Dairy

technique for predicting the dry matter digest Sci 62 807 1979

ibilit of grasses JSci Food Agric 24 J1 21l) Sncdccor G Cochran WG Statistical

1426 197J lelhods 9th printing Sixth edition 91119

15) Kane EA Ely RE Jacobson WC loore LA Tre Iowa State University Press Ames Iowa

Comparison of alous digestion triill tech CS 1978

niques ith dain cttle J Daif 36 320 Valdes EV Iunte RB Jones GE Comparison

333 1953 of iear infrared calibration for estimating ill 16) Leaer 1) Cam plilg RC 1 101nte Tle effect lilrrJ digestiiJility in whoie plant corn hybrics

len1 fleding on lhl digestibilit of diets lor Can I Anim Sci 6 57562 1987 sheep and cattle Anim PIod II bull 11middot18 1969 30 an PJ Labora~or methods for cvaludt

cLeed 1 linson IJ The aCClll2C of pr( the energy value or leeds(llffs in

dieting dry mlttPr digestibilit of gtasses energ sources for liestock Swan II and

from lignir analss b different Lewis f) 839middot1 Butterworth amp Co Ltd

lllethods J Sci Food -1ric 2S 91 L 19~L 1~171

8(i(j

ITRtOOcDL K[1RIILH ISIIIDA SIlIl3ATA cBE and KAt1EOKA

strains of rye grass at middotaryin stages 0 r11C1tu 18) Jvlorimoto II Dobutsu eiyo shikenho (Experi

rity J Agric Sci Camb 62 399 1961 mental method of animal nutrition 280

6) Barnes RJ ear infra rcd spectra of ammona 119 Yokendo Tokyo 197L (in )apancs()

treated straw and isolated cell walls nim 191 -lurray L Williams Pc Chemical principles of

Feed Technol 21 209211)1988 near infrared technology in near infrared

7) Bondy AA nimal nutrition 3)~ icy technology in agricultural and food industries

Intersciencc PublicaticJr) John Wiley 6 Sons Williams Pc Norris KH 3031 American asso

Ltd Chichester lew York BrisiJane Toronto ciation of cereai chemists Inc SL Paul l1inne

Singapore 1987 tS1990

8) Coelho 1 lembry FG Barton FE Saxton 1 20) National Research Council of Agriculture For

A cornparison of rtdcrobia enrymltlcic chemi estry and Fisheries Japanese Feeding Stand

cal and nearinfrared reflectance spectroscopy ani fer Dairy Catlie 66-68 NRCAFF Ministry

methods in forage ealuation nm Feed Sci of Agriculture Forestry anc Fisheries 1987

TechtlD 20 2192]1 1988 21) aaratne IlYRG Ibrahim lll Schiere JB

9) De Boccr JL Cocn BG anacker JM Comparison of four techniques for predicting

Boucque Ch The use of nirs t) prwlic the digestibilit of ropical feeds Anim Feed

chemical composition and the energy liue Techno 29 bull 20921 1990

compound feeds for catllc nim Feed Sci 22) [on sOllchi koukai Soushiryou no

Techno 51 243-251 1995 hinshilsll hyouka gaid()~)Ukkujikuli shiryou

10) Elam CJ Dinis RE Lignin excretion [l cCluie hinshitsll houka kcnkyuukiiher~ 1

fed a mixed ~atjon J Anim Sci ~Q 4il(j KyoUa Toko 1995 In Japanese)

1961 0orris KII Barnes RF loorc JE Shen S

11) El RE Kane E Jacobson ( loore LA Predicting forage quaE by infrared

Studico the compOoilic)l of Ignin i~oiaed ctance spectroscopy 1 Anim Sci 43 887-889 from orchard grass hay Cll at four stages of 1976

maturit and frum the corre~polldlllg cces ) 211 Orskm ER Protein nutrition in ruminants 2 Dai Sci 36 3middot6-lJi nel edition 89 157 Academic Press Ltd

12) Givens D1 Baker C dalllson I CjSS R Londo] 1992

Inlluence of gn)wltl lype and on the 25) Osborne EG Fearn T Ncar infrared spectros

prediction of the Illcabolisablc conent copy in food analysis 3740 133 Longman

of herbage by nearinflued rellectance spec Scientific amp TechnicaL Longman Group CK

troscopy nim Feed Sci Techno 37 281 Limited Ergland 1986

295 1992 26) Penl1ing PD Johnson RIL Th of internal

13) lolechek JL Shenk JS YaH2 1 Arthun [) markers to estirna~e herbage digestibility and

Prediction of forage qualit using ncar infra intake I Potentiaily indigestible cellulose and

red rellectanc( spectroscopy on (sophageal insoluble ash J cgric Sci Camb 100

fistula samples from cattle or mountain r2nge 127Ll1 1983 J Anim Sci 55 971 97S 1982 Shenk JS Westerhaus 0 Hoover MR Anal y

14) Jones DIH Hayward IV A cellulase digestion sis of by infrared relleclance J Dairy

technique for predicting the dry matter digest Sci 62 807 1979

ibilit of grasses JSci Food Agric 24 J1 21l) Sncdccor G Cochran WG Statistical

1426 197J lelhods 9th printing Sixth edition 91119

15) Kane EA Ely RE Jacobson WC loore LA Tre Iowa State University Press Ames Iowa

Comparison of alous digestion triill tech CS 1978

niques ith dain cttle J Daif 36 320 Valdes EV Iunte RB Jones GE Comparison

333 1953 of iear infrared calibration for estimating ill 16) Leaer 1) Cam plilg RC 1 101nte Tle effect lilrrJ digestiiJility in whoie plant corn hybrics

len1 fleding on lhl digestibilit of diets lor Can I Anim Sci 6 57562 1987 sheep and cattle Anim PIod II bull 11middot18 1969 30 an PJ Labora~or methods for cvaludt

cLeed 1 linson IJ The aCClll2C of pr( the energy value or leeds(llffs in

dieting dry mlttPr digestibilit of gtasses energ sources for liestock Swan II and

from lignir analss b different Lewis f) 839middot1 Butterworth amp Co Ltd

lllethods J Sci Food -1ric 2S 91 L 19~L 1~171

8(i(j