Challenges in measuring moisture content of feeds

N. Thiex and C. R. Richardson

J Anim Sci 2003. 81:3255-3266.

The online version of this article, along with updated information and services, is located on
the World Wide Web at:
http://jas.fass.org/cgi/content/full/81/12/3255

www.asas.org

Downloaded from jas.fass.org by on October 6, 2010.

 Challenges in measuring moisture content of feeds

N. Thiex*1 and C. R. Richardson†2

*South Dakota State University, Brookings 57007; †Texas Tech University, Lubbock 79409

ABSTRACT: Accurate determination of the moisture methods for the evaluation of empirical methods. The
(water) content in individual feed ingredients and objective of this paper is to review methods for de-
mixed feeds is critical throughout the feed industry. termining moisture, review comparisons among mois-
Most analytical methods used to determine apparent ture methods for various feedstuffs, make recommenda-
water content of feedstuffs are empirical, estimating tions for a reference method, and make general recom-
water by evaporation and loss of weight on drying (oven mendations toward improving the results of moisture
drying methods). These methods differ greatly in effec- testing. The need to evaluate and improve moisture
tiveness, resulting in bias. Bias associated with mea- methods and standardize practices in laboratories is
suring the water content of feedstuffs is a concern not evident from this study. It also is evident that the meth-
only because of the lack of confidence in the moisture ods appropriate for a specific feed ingredient or feed
value itself, but also because moisture determinations should not be extended to all feeds without proper vali-
affect accurate quantification and expression of other dation to the new matrices. Part of the validation for
nutrient values. Methods for determining moisture in empirical methods should be comparison to Karl Fis-
feeds have frequently been borrowed from the cereal, cher or other the direct methods. It also is recommended
forage, or other applications without validating the ex- that the results obtained using oven methods not be
tension of the method. Methods such as Karl Fischer termed “moisture;” rather, they should be termed “loss
titration measure water by direct comparison to a cali- on drying,” and the drying conditions should become
bration standard for water and can be used as reference part of the term.

Key Words: Analytical Methods, Drying Methods, Feeds, Moisture Content, Ovens

2003 American Society of Animal Science. All rights reserved. J. Anim. Sci. 2003. 81:3255–3266

Introduction vitamins in diets for animals. Moisture determinations

are used to convert all other nutrients to a dry matter
Perhaps no analysis is more widely used in the ag- basis; therefore, errors in moisture determination are
ricultural sector, and no more widely abused than that incorporated into other nutrient concentration calcula-
of moisture. This issue prompted a presentation and tions; 5) proper moisture concentrations in diets are
discussion at the Contemporary and Emerging Issues necessary for optimum intake and performance of ani-
Symposium at the 2002 ADSA/ASAS/CSAS Joint Con- mals; thus, an accurate determination of moisture in
ference. ingredients is necessary.
Accurate moisture determinations are critical in the Factors affecting accurate determination include
feed industry for many reasons: 1) water is weight and range of moisture content, sampling of feedstuffs, trans-
must be paid for when grain [feed] is bought and sold port and storage of laboratory samples, laboratory sam-
(Hunt and Pixton, 1974); 2) water is weight and must ple preparation, including grinding, and bias and vari-
be shipped; 3) moisture content plays a role in storage ability associated with the specific analytical method
conditions; 4) moisture in feedstuffs and mixed feeds used. Inaccurate moisture analysis has a direct and
works as a diluent to energy, protein, minerals, and negative influence on the precision of diet formulation,
in balancing consumption of other nutrients, and in
predicting performance.
Feedstuffs have been classified by Kellems and
1
Correspondence: Box 2170, ASC 133 (phone: 605-688-5466; fax: Church (2002) as including eight categories: dry rough-
605-688-6295; E-mail: [email protected]). ages, pasture and range grasses, ensiled roughages,
2
Center for Feed Industry Research and Education, Dept. of Ani-
mal and Food Sciences, Box 42141, Texas Tech Univ.
high energy concentrates, protein sources, minerals,
Received November 26, 2002. vitamins and additives. An animal feed may contain
Accepted March 31, 2003. one or all eight categories, complicating the analytical

3255

Downloaded from jas.fass.org by on October 6, 2010.

3256 Thiex and Richardson

matrix. Methods for determining moisture in these ma- Materials and Methods
trices are compared in this study.
Moisture analysis data are presented from sources
Analytical Methods that compare various loss-on-drying (LOD), commonly
referred to as oven drying, methods to the Karl Fisher
The methods in routine use to determine moisture method. Statistical treatments varied among the au-
in feedstuffs are “loss on drying” or “oven-drying” which thors. The reliability of the NIR technique for feeds was
estimate moisture by evaporation. These are empirical evaluated using both oven drying and the Karl Fisher
methods wherein the moisture results obtained are de- method as the primary methods for NIR calibrations.
fined by the drying conditions (time, temperature, etc.).
Problems with this approach have been studied and Analytical Methods
documented by Mo and Tjornhom (1978). During oven A large number of moisture methods are discussed or
drying, volatile substances other than water are lost compared throughout this paper. A letter designation,
and side chemical reactions occur during the heating brief description and the reference for each are provided
process. Windham et al. (1987) and Thiex and Van in Table 1. The methods will be referred to by the letter
Erem (1999) reinforced Mo and Tjornhom’s findings, designation in subsequent tables.
suggesting no progress on improvement of moisture The Karl Fischer method used by Thiex and Van
testing over two or more decades. Even though oven Erem (1998) was approved as first action AOAC Official
methods are prone to error, they remain commonly used Method 2001.12, Determination of Water (Moisture)/
because the determinations are fast and inexpensive Dry Matter in Animal Feed, Grain and Forage (Plant
to perform. Tissue), Karl Fischer Titration Methods, based on an
A second type of analytical method to determine mois- interlaboratory collaborative study published in 2002
ture methods are those that extract the water molecules (Method KF1) (Thiex and Van Erem, 2002). The princi-
from feedstuffs and measure water concentration ple of this method is to extract water from animal feed
against a calibration standard. This includes the Karl or forage material into methanol–formamide (1 + 1)
Fischer (KF) method, which was collaboratively studied directly in the Karl Fischer titration vessel using high-
and received first action status as AOAC Official speed homogenization. The water is then titrated at
Method 2001.12 in 2001 (Thiex and Van Erem, 2002). It 50°C with one-component Karl Fischer reagent based
might also include azeotropic distillation (AOAC, 2000) on imidazole. The high-speed homogenization ruptures
however, this method is no longer widely used in labora- cells, allowing for titration of free and bound water.
tories.
A third approach for determining moisture is the de- Comparison from AAFCO Check Sample Program
velopment of NIR calibrations based on a primary
method. The approach was established by Windham et The American Association of Feed Control Officials
al. (1991) as an alternative to oven methods for forages. (AAFCO) operates the AAFCO Check Sample program
A critical aspect of NIR methods is in the choice of the for feed laboratories. The program consists of monthly
primary method for calibration. proficiency samples sent to ∼300 participating labora-
tories worldwide. The series includes a variety of com-
plete feeds, premixes, concentrates and supplements
Sources of Error
with drugs, antibiotics, minerals, and vitamins at levels
Hunt (1974) described moisture behavior and mea- normally encountered in commercial products. Ana-
surements related to cereal grains and discussed lysts are asked to perform single determinations on the
sources of error. Horwitz (1990) elaborated on sources check sample on two separate days and submit the
of error specific to the determination of moisture by results, and code data to specific analytical methods.
evaporation. Sources of error that apply to all moisture Moisture data was extracted from 20 years of check
methods include: 1) representative-ness of laboratory sample data. Five oven methods, a Karl Fischer
sample; 2) storage conditions of both laboratory and (Method G), azeotropic distillation (Method H), and a
analytical samples; 3) grinding techniques (exposure to thermogravimetric (Method F) method were compared
air, generation of heat, contamination, fineness of to oven drying under vacuum at 95°C for 5 h (Method
grind, necessity for two step moisture determination J). The oven methods included drying at 135°C for 2 h
for high moisture feeds); 4) weighing errors; 5) test (Method A), drying at 104°C for 3 h (Method B), drying
portion size; 6) room humidity; 7) nonaqueous losses under vacuum at 60°C for 18 h (Method C), drying at
or interferences (specificity/selectivity of the method). 102°C for 16 h (Method D) and drying under vacuum
Additional sources of error for oven drying methods at 70°C (Method E).
include: 8) drying time and temperature; 9) oven (tem-
Comparisons of Oven Method to Karl Fischer
perature stability, heating uniformity, ventilation, air-
for Forages
flow, recovery rate, thermometer accuracy); 10) drying
pan (area, nature, size and placement in oven); 11) des- Two published studies compared oven methods to
iccant. Karl Fischer for forages. The first was that of Windham

Downloaded from jas.fass.org by on October 6, 2010.

 Measuring moisture content of feeds 3257
Table 1. Analytical methods for moisture included in method comparisons
Method Reference/Association Matrix validated Method description

A AOACa 930.15, NFTAb 2.1.1 Some feeds Oven, 135°C, 2h

B AOAC 935.29 Malt Oven, 104°C or 105°C, 3h
C AOAC 925.45 Sugar Vacuum oven <50mm Hg, 60°C, 18h
D AOAC 950.46c Meat Oven, 102°C, 16hc
E AOAC 31.007d High-sugar molasses, liquid, semi-liquid Vacuum oven =20mm Hg, 70°C
F Manufacturer’s method (Leco) Thermogravimetic analysise
G AOAC 966.20 Liquid molasses Karl Fischer
H AOAC 925.04 Feeds Azeotropic distillation
I Experimental Oven, 104°C or 105°C, 6h
J AOAC 934.01 Feed Vacuum oven = 100 mm Hg, 95°C, 5h
KF1 AOAC 2001.12 Feed/forage Karl Fischer, homogenization extraction
KF2 Robertson and Windham (1983) Forages Karl Fischer, ball mill extraction
L Experimental Oven, 110°C, 3h
M Experimental Oven, 110°C, 15h
N AOAC 925.45 modified Vacuum oven 60 mm Hg, 60°C, 20h
O AOAC 934.01 High-molasses feed Vacuum oven = 30 mm Hg, 70°C, 20h
P Corn Refiners Assn. G-16f Vacuum, 100°C, 4h
a
AOAC methods, unless otherwise indicated, are available in: AOAC INTERNATIONAL. 2000. Official Methods of Analysis (OMA). 17th
ed. AOAC INTERNATIONAL, Gaithersburg, MD.
b
NFTA methods are available at http://www.foragetesting.org or in: Undersander, D., D. R. Martens, and N. Thiex. 1993. Forage Analysis
Procedures. National Forage Testing Association, Omaha, NE.
c
This references three methods—any one of whose results could be grouped in this total. Section A is equivalent to OMA 934.01; Section
B has two subsections: (a) air-dried for 16-18 h at 100-102°C; (b) at 125°C to constant weight.
d
Association of Official Analytical Chemists, Inc. 1984. Official Methods of Analysis. 14th ed. Association of Official Analytical Chemists,
Inc., Arlington, VA.
e
Since the instrument is programmable, the laboratory may have used either the 104°C or 135°C temperature; however, the data suggest
that only the 104°C data are being reported.
f
Available from the Corn Refiners Assoc., Washington, DC.

(1987) and the second was that of Thiex and Van and Grassland Council, the National Hay Association
Erem (1999). and forage testing laboratories in a concentrated effort
Thirty forages evaluated by Windham et al. included to improve the accuracy of forage testing and build
six bermudagrasses, six temperate grasses, six le- grower confidence in testing animal feeds. One of the
gumes, six silages, and six silage based rations. All activities of the organization to improve repeatability
materials were dried by forced air oven at 65°C, and and accuracy of forage testing among laboratories is a
ground in a Wiley mill to pass through a 1-mm screen. laboratory proficiency testing program. Bimonthly sam-
Moisture was determined by oven drying at 135°C for ples are sent to participating laboratories. The series
2 h (Method A) and by Karl Fischer titration (Method includes three alfalfa hay, one alfalfa-grass mixed hay,
KF2) using the extraction procedure of Robertson and one grass hay and one corn silage annually.
Windham (1983) and water calculation as reported by Laboratories are evaluated on performance compared
Jones (1985). to a reference method. The NFTA reference method for
The forages evaluated by Thiex and Van Erem (1999) moisture was established in 1993 as oven drying at
included hay, haylage, and corn silage. Hay materials 135°C for 2 h (Method A) (Undersander et al., 1993).
included four each of grass hay, legume hay, and le- In 2000, the NFTA Board of Directors established a
gume-grass mix. Haylage materials included one grass moisture task force to investigate the reproducibility,
silage, nine legume silages, and one legume/grass accuracy, and applicability of various oven moisture
mixed silage and ten corn silages. Materials were dried methods for estimating moisture in forages. Two alfalfa
in a microwave oven, and ground allowing to pass a hay, two grass hay and two corn silage materials were
1mm sieve in a cyclone mill. sent to each of eight labs. All materials came from previ-
Moisture (as water) was determined in triplicate on ous NFTA Proficiency Testing materials. Each lab ran
all materials using the Karl Fischer method, and in moisture by three oven methods: 105°C for 3 h (Method
duplicate with the oven methods as follows: drying at B); 105°C for 6 h (Method I); and 135°C for 2 h (Method
135°C for 2 h (Method A), 104°C for 3 h (Method B), A). One lab ran Karl Fischer (Method KF1).
and 104°C for 6 h (Method I).
Comparisons of Oven Methods to Karl Fischer
National Forage Testing Association Assessment for Nonurea and Urea Feeds
of Moisture Methods in Forages
Six feeds containing urea and ten feeds containing
The National Forage Testing Association (NFTA) was no urea were compared by Thiex and Van Erem (1999).
founded in 1984 as a joint effort of the American Forage The feed materials were ground to pass a 1mm sieve

Downloaded from jas.fass.org by on October 6, 2010.

3258 Thiex and Richardson

in a Retsch centrifugal mill, except for the soybeans laboratories also had an early interest in the Karl Fis-
and cat food, which were ground using a Tecator Kni- cher method and first adopted it for the determination
fetec mill. Urea containing feeds represented a range of of water in concentrated steep water, which contains
urea levels, from 0.08 to pure feed grade urea. Nonurea water-soluble extractives from the steeping of corn. For
containing feeds represented a cross section of commer- steepwater, azeotrope distillation was not practical in
cial feeds and feed ingredients. Moisture (as water) was routine work and oven-drying methods were not repro-
determined in triplicate on all materials using Karl ducible owing to variable contents of volatile matter
Fischer (Method KF1), and with oven methods as fol- other than water in the product. In 1996, a Karl Fischer
lows: drying at 135°C for 2 h (Method A), 104°C for 3 method was proposed by Corn Products International
h (Method B), 95°C for 5 h under vacuum (Method J), to measure the moisture contents of whole corn based
104°C for 6 h (Method I), 110°C for 3 h (Method L). upon the extraction of water by ball milling a test sam-
In another study, Shreve et al. (2000) compared four ple to a paste in the presence of anhydrous methanol.
drying methods with Karl Fischer moisture titration It was subjected to an inter-laboratory evaluation and
(Method KF1) on ten urea supplements and feed grade was found to agree with results obtained by azeotrope
urea to investigate method effect on urea loss. Supple- distillation (Bernetti et al., 1996). The CRA therefore
ments ranged from 35.4% to 71.6% crude protein and became interested in the method of Thiex and Van Erem
from 3.2% to 49.3% nonprotein nitrogen. All moisture (1998) (Method KF1). A technical committee of the CRA
determinations were made in triplicate. Drying meth- decided that it could be adopted for corn, but it would
ods included: convection oven at 105°C for 3 h (Method be necessary to obtain supplementary evidence of appli-
B), vacuum oven 60 mm Hg at 60°C for 20 h (Method cability to corn gluten meal and corn gluten feed before
N), vacuum oven 60 mm Hg at 95°C for 5 h (Method adoption as a CRA method. The CRA objective was to
determine if Karl Fischer Method KF1 could be estab-
J), and vacuum oven 30 mm Hg at 70°C for 20 h
lished as a new primary method for standardizing and
(Method O). Means were tested at the α = 0.01 level
monitoring the secondary loss-on-drying methods in
with a paired t-test.
corn and corn products. (Raffaele Bernetti and Jennifer
Snyder, unpublished data).
Comparison of Oven Methods to Karl Fischer Method The CRA committee chose samples available from
for Steam-Flaked Corn and Corn Products internal check studies. Typically, these samples are
used to monitor product quality between plants at each
Five steam-flaked corn batches produced from two
company and, therefore have known analytical histor-
different steam flaking units at Texas Tech University
ies and traceability. Two samples of corn gluten meal
were used to determine moisture addition as compared
were submitted by Corn Products International and
to whole shelled corn of the same source, and moisture
samples, respectively, of corn gluten feed and corn glu-
level of the steam-flaked batches as determined by oven
ten meal were submitted by Roquette America for anal-
drying and Karl Fischer. A small scale pilot steam
ysis by Karl Fischer (Method KF1). All four samples
flaker and a commercial model steam flaker were used had been analyzed by oven method P.
to compare moisture levels of corn processed under rou-
tine conditions. Flaking units differed in the design Assessing the Potential for NIR Moisture
of the steam cabinets (round vs. rectangular), but the Determinations in Feed
diameter and corrugation of the rolls were the same.
Moisture was determined in triplicate on all steam- Windham et al. (1987, 1991) had established NIR as
a valid technique for moisture determination in forages.
flaked samples and the whole shelled corn sample by
The potential for extending an NIR method for moisture
oven drying overnight (15 h) at 110°C (Method M), and
in forages to determine moisture in feeds was reported
by Karl Fischer (Method KF1). Oven drying was
by Thiex and Van Erem (1999). Evaluation was based
achieved by weighing samples into aluminum pans and
on a standard error of calibration (SEC), correlation
placing them inside a forced air oven without vacuum coefficient, and partial least squares calibration.
overnight. Samples were removed from the oven, placed Partial least squares calibration equations were gen-
in a desiccator until they were reweighed to determine erated using a 1, 4, 4, 1 math treatment, where the first
loss in weight. value indicates the first derivative, the second value
The Corn Refiners Association (CRA) is experienced indicates a gap of four over which the derivative is
in the measurement of moisture of corn raw material calculated, the third value is the number representing
and the products of the corn wet milling process. By the smoothing of points, and the fourth value indicates
the mid 1950’s, CRA industry members had come to rely no second smooth.
on azeotrope distillation of water-toluene and water-
benzene systems as methods of reference to standardize Results and Discussion
oven drying methods (Analytical Methods of the mem-
ber companies of the Corn Refiners Association Inc,
Comparisons of Methods Reported to the AAFCO
Check Sample Program
1999), and had standardized two oven methods relative
to azeotrope distillation: 80°C, 20 h under vacuum and The comparison of check sample results submitted
100°C, 4 h under vacuum (Method P). CRA member by participating laboratories over a 20-year period on

Downloaded from jas.fass.org by on October 6, 2010.

 Measuring moisture content of feeds 3259
eight methods for the determination of moisture is pro-

92.54 – 109.03
9.78 – 22.53
Distillation
vided in Table 2 (George W. Latimer, AAFCO Check
Sample Committee Chair, unpublished data). These

H
comparisons are of special interest because they repre-

100.78
101.41

119.27
13.64
69.42

13.54
Direct methods
sent the methods routinely practiced in feed labora-

13
tories in the US and other countries. The most widely
used methods are Method J (at least 223 check sam-

Table 2. Moisture results of loss on drying and direct methods for AAFCO check sample data analyzed over a 20-y period,

88.62 – 95.52
11.66 – 16.63
ples), Method B (223 check samples), followed by
Method A (157 check samples) and Method D (96 check

KF
G
samples). Recoveries were based on AOAC Method

153.39
92.07
94.00
13.71
53.81

14.88
934.01 (Method J). Mean recoveries of water range from

63
92.07% for Method G to 113.17% for Method A. How-
ever, averaging over all samples masks the extent of the

92.72 – 107.24
problem. Minimum and maximum percent recoveries

7.34 – 18.24
within a method for individual check samples varied

TGA
widely, and were 100.10 and 202.03, respectively, for

F
Method A; 76.45 and 155.04 for Method B; 61.28 and

100.13

115.79
99.98

10.82
81.35

10.82
11
331.92 for Method C; 79.09 and 153.66 for Method D;
75.45 and 197.05 for Method E; 81.35 and 115.79 for
Method F; 53.81 and 153.39 for Method G; and 69.42

Compared with moisture results obtained using vacuum at 95 to 100°C for 5 h Official Methods of Analysis (OMA) 934.01.
75.15 – 47.78
expressed as a percentage of Method J (Table 1)a,b
and 119.27 for Method H. Few conclusions can be drawn

25.30 – 6.47
vacuum
from this comparison about the merits of a particular

70°C
E
method. But the wild fluctuations demonstrate the in-

111.46
100.93

197.05
39.27
75.45

35.23
appropriateness of extending an empirical oven drying

7
method for estimating moisture to matrices for which
they have not been validated.

97.97 – 101.73
8.12 – 10.81
102°C/16 h
Comparisons of Oven Method to Karl Fischer
for Forages
Loss on drying methods

153.66
99.85
98.74

79.09
9.28

9.29
Percent moisture bias in dry forage materials of oven 96

methods compared to Karl Fischer are illustrated in

Figure 1a and average percent recovery for oven meth-

79.28 – 133.58
ods compared to Karl Fischer methods are illustrated

42.56 – 83.30
vacuum/18 h

in Figure 1b. Windham et al. (1987) reported Karl Fis-

60°C

cher and oven means (Method A) different (P < 0.05)

C

106.45

331.92
97.18
56.32
61.28

52.92

for legume and temperate forage, and no difference for

19

bermudagrass. Likewise, Thiex and Van Erem (1999)

reported Oven Method A different (P < 0.05) from Karl
AAFCO = Association of American Feed Control Officials, Inc.
Fischer for seven of the twelve hay materials. Both
97.66 – 99.32
5.75 – 6.93

authors observed a 121% recovery of moistures from

104°C/3 h

legume hay for Oven Method A compared to Karl Fis-

B

cher methods. Windham et al. observed a 109% recovery

155.04
98.49
98.38

76.45
6.28

6.38

RSD = relative standard deviation = (SD/x) × 100.

223

for temperate grasses, and a 100% recovery for bermu-

dagrass for Oven Method A. Thiex and Van Erem (1999)
observed a 111% recovery for legume/grass mixed hay
110.35 – 115.99

and a 108% recovery for grass hay for Oven Method

15.50 – 20.49
135°C/2 h

A compared to Karl Fischer. Observations of the two

A

authors were very similar, even though they were using

Adapted from Latimer (2002).
17.71

15.65
113.17
107.14

100.10
202.03

different extractions prior to the Karl Fischer titration.

Number of check samples.
151

Both authors concluded that Oven Method A was inap-

propriate for use with legume and temperate forage ma-
terials.
Median, % recoveryd
Standard deviation

Thiex and Van Erem (1999) reported on two addi-

Max, % recovered
Min, % recovered

Confidence limits

tional oven methods: Oven Method B means were differ-

x, % recoveryd

ent (P < 0.05) from the Karl Fischer mean for just one
of twelve hay materials and Oven Method I mean differ-
variable
Method

ent (P < 0.05) from the Karl Fischer mean for two of
RSDe

SD
No.c

d
a
b

e
c
x

twelve hay materials. They observed an average per-

Downloaded from jas.fass.org by on October 6, 2010.

3260 Thiex and Richardson

Figure 1. Forage moisture bias of oven methods compared to Karl Fischer (a). Mean percentage of forage moisture
recovery for oven methods based on Karl Fischer (b).

cent recovery for Oven Method B compared to Karl to Karl Fischer of 100% for legume hay, 93% for legume/
Fischer of 99% for legume hay, 93% for legume/grass grass mixed hay, and 99% for grass hay.
mixed hay, and 96% for grass hay. They observed an Windham et al (1987) reported Karl Fischer and oven
average percent recovery for Oven Method I compared means (Method A) different (P < 0.05) for silages and

Downloaded from jas.fass.org by on October 6, 2010.

 Measuring moisture content of feeds 3261
Table 3. Comparison (correlation coefficient and slope) of Karl Fischer moisture values,
and moisture values obtained using oven methods KF1 = Karl Fischer,
Method A = 135°C/2 h, Method B = 104°C/2 h, Method I = 104°C/6 h,
Method J = 95°C vacuum/5 h, Method L = 110°C/3 ha
KF1 vs. A KF1 vs. B KF1 vs. I KF1 vs. J KF1 vs. L

Material r Slope r Slope r Slope r Slope r Slope

Hay 0.85 0.76 0.97 0.95 0.97 0.92

Haylage 0.50 0.30 0.45 0.41 0.42 0.38
Corn silage 0.61 0.62 0.84 0.79 0.78 0.74
Nonurea feed 0.30 0.86 0.99 1.08 0.99 1.15 0.95 1.00 0.98 1.19
Urea feed 0.44 −0.30 0.96 1.11 0.75 1.03 0.30 0.80 0.76 1.09
a
Adapted from Thiex and Van Erem (1999).

different (P < 0.05) for silage-based rations. Likewise, Method A overestimated moisture in hay. All oven
Thiex and Van Erem (1999) reported Oven Method A methods seriously overestimated moisture in haylage.
different (P < 0.05) from Karl Fischer for ten of eleven Likewise, oven methods overestimated moisture in corn
haylage materials and for all ten corn silage materials. silage, however, Method A overestimated moisture
Windham et al (1987) observed a 151% recovery for most dramatically.
Oven Method A compared to Karl Fischer for silage and For ensiled forages, the loss of volatile substances
a 121% recovery for silage based rations. Thiex and other than water is drastic with oven heating. This is
Van Erem (1999) observed a 159% recovery for legume logical due to the loss of volatile fatty acids that occur
silage, a 191% recovery for mixed haylage, 159% recov- in ensiled products. Overall, the Method B oven method
ery for grass silage and 133% recovery for corn silage. most closely approximated Karl Fischer. It appears that
When averaged together, the recovery of 148% on all with this method only 96% of the water is removed;
silage materials compares to the 151% recovery ob- however, the loss of volatile substances other than wa-
served by Windham et al. ter more than compensates for the incomplete removal
Thiex and Van Erem (1999) reported on two addi- of water from the ensiled products.
tional oven methods for silage materials: Oven Method
B means were different (P < 0.05) from the Karl Fischer National Forage Testing Association Assessment
mean for nine of eleven haylage materials and five of of Moisture Methods in Forages
ten corn silage materials. They observed an average
percent recovery for Oven Method B compared to Karl The average moisture for the Karl Fischer, Method
Fischer of 120% for alfalfa haylage, 139% for mixed A, Method B, and Method I methods were 7.08, 8.93,
haylage, 123% for grass silage, and 113% for corn silage. 7.35, and 7.71% respectively (NFTA Moisture Task
Oven Method I means were different (P < 0.05) from Force Reports, 2001). Standard deviations for the re-
the Karl Fischer mean for ten of eleven haylage materi- spective methods were 0.73, 0.81, 0.82, and 0.80. Per-
als, and for seven of ten corn silage materials. They cent moisture bias as compared to Karl Fischer is illus-
observed an average percent recovery for Oven Method trated in Figure 1a.
I compared to Karl Fischer of 127% for alfalfa haylage, Analysis of variance for each sample revealed that
148% for mixed haylage, 130% for grass silage, and labs and methods all resulted in different oven moisture
117% for corn silage. values. All labs’ results were within ±3 standard devia-
Thiex and Van Erem (1999) reported correlation coef- tions of the mean for each sample. For one of the grass
ficients and slope of Karl Fischer and each oven method hay samples, removing two outlying labs eliminated
for forage materials (Table 3). The correlation coeffi- the significant effect of lab on oven moisture results. In
cients (r) for hay, haylage, and corn silage for Karl every case, Method A produced higher moisture results
Fischer vs. Method A were 0.85, 0.50, and 0.61, and than the other methods. Little difference was observed
the slopes were 0.76, 0.30, and 0.62, respectively. The in reproducibility among the various methods.
correlation coefficients for hay, haylage, and corn silage The NFTA study grouped all forages (hays and si-
for Karl Fischer vs. Method B were 0.97, 0.45, and 0.84, lages) together for statistical analysis. Paired t-tests
and the slopes were 0.95, 0.41, and 0.79, respectively. comparing each oven method to Karl Fischer moisture
The correlation coefficients for hay, haylage, and corn indicated no difference (P < 0.01) between Method B
silage for Karl Fischer vs. the Method I method were and Karl Fischer values for the materials studied. Both
0.97, 0.42, and 0.78, and the slopes were 0.92, 0.38, and the Method I and Method A resulted in higher moisture
0.74, respectively. results than Karl Fischer (Method KF1).
On the average for hay, Methods B and I approxi- The NFTA Moisture Committee concluded that,
mated the values obtained by Karl Fischer, whereas based upon agreement with Karl Fischer, Method B

Downloaded from jas.fass.org by on October 6, 2010.

3262 Thiex and Richardson

Table 4. Moisture content and percent recovery of different feed and feed ingredients
determined by Karl Fischer and five oven methodsa
Method KF1 A B J I L
Feedstuff or mixed feed material Karl Fischer 135°C/2 h 104°C/3 h 95°C vacuum/5 h 104°C/6 h 110°C/3 h

Mean moisture, %
Nonurea
Number of materials where method
differs significantly (P < 0.05)
from the Karl Fischer method 7/10 10/10 10/10 9/10 7/10
Milk replacer 4.19 10.99b (262) 3.68b (88) 2.64b (63) 4.14 (99) 4.22 (101)
Corn, shelled 12.34 11.82b (96) 11.46b (93) 10.43b (85) 10.90b (88) 10.44 (85)
Cat food 8.39 8.23 (98) 7.23b (86) 7.00b (83) 7.63b (91) 7.96b (95)
Hog feed 11.81 11.42b (97) 10.84b (92) 10.62b (90) 10.78b (91) 10.94b (93)
Soybeans 12.58 11.61b (92) 10.9b (87) 10.45b (83) 11.47b (91) 11.71b (93)
Barley, hulless 10.78 9.63b (89) 9.32b (86) 9.60b (89) 9.09b (84) 9.60b (89)
Meat and bone meal 6.77 6.79 (100) 5.98b (88) 5.94b (88) 6.06b (90) 6.61 (98)
Swine feed with oxytetracycline 7.61 8.70b (114) 6.79b (89) 6.70b (88) 7.03b (92) 7.39 (97)
Pasture supplement 9.41 9.59 (102) 8.35b (89) 8.46b (90) 8.46b (90) 8.84b (94)
Swine feed with carbadox 7.61 7.94b (104) 6.57b (86) 5.73b (75) 6.64b (87) 7.12b (94)
Containing urea
Number of materials where method
differs significantly (P < 0.05)
from the Karl Fischer method 6/6 2/6 5/6 4/6 4/6
Lamb grower (0.81% urea) 9.96 10.92b (110) 9.60b (96) 9.43b (95) 9.36b (94) 9.09b (91)
Forage block (3.3% urea) 11.18 12.63b (113) 11.28 (101) 10.39b (93) 11.43 (102) 11.25 (101)
Custom mix (6.6% urea) 6.83 9.28b (136) 6.79 (99) 7.18 (105) 7.14 (105) 6.86 (100)
Beef feedlot supplement (10.0% urea) 6.45 11.26b (175) 7.14b (111) 8.47b (131) 7.83b (121) 7.26b (113)
Meal supplement (∼37% urea) 8.89 23.16b (261) 10.40 (117) 14.39b (162) 13.05b (147) 12.41b (140)
Pure feed-grade urea 0.193 30.26b (15679) 1.76 (912) 7.29b (3777) 3.85b (1995) 3.87b (2005)
a
Adapted from Thiex and Van Erem (1999). N = 3 for Karl Fischer and N = 2 for oven methods. Percent recovery compared to Karl Fischer
in parenthesis.
b
Means within a feed row with a superscript differ (P < 0.05) from the Karl Fischer mean.

most closely represents the true moisture content of Correlation coefficients (r) and the slope of the Karl
alfalfa and grass hays. Even for corn silage, this method Fischer and each oven method are reported for feed
appears better than other oven methods currently materials in Table 3. The correlation coefficients for
available. Their conclusion is consistent with those of feed with urea and feed without urea for Karl Fischer
Thiex and Van Erem (1999) and Windham et al. (1987). vs. the Method A method were 0.44 and 0.30, and the
slopes were −0.30 and 0.86, respectively, indicating
Comparisons of Oven Methods to Karl Fischer Method A is unsuitable to estimate water in either
for Nonurea and Urea Feeds group. The correlation coefficients for feed with urea
and feed without urea for Karl Fischer vs. the Method
The average percent moisture obtained for feedstuff
B were 0.96 and 0.99, and the slopes were 1.11 and 1.08,
and mixed feed materials and average percent recovery
respectively, indicating Method B may be a suitable
of moisture compared to the Karl Fischer method are
method to estimate water in feed. The correlation coef-
reported in Table 4. Also provided is a summary of the
ficients for feed with urea and feed without urea for
number of materials which were different (P < 0.05)
Karl Fischer vs. the Method I were 0.75 and 0.99, and
from the Karl Fischer means for each oven method. The
percent recovery of water for the Method A, Method B, the slopes were 1.03 and 1.15, respectively, indicating
Method J, Method I, and Method L means were 116, little or no improvement in performance over Method
88, 83, 90, and 94 respectively, for nonurea feeds. B. The correlation coefficients for feed with urea and
For urea feed materials, percent recovery of water feed without urea for Karl Fischer vs. the Method J
increases exponentially as the urea concentration in- were 0.30 and 0.95, and the slopes were 0.80 and 1.00,
creases, (See Figure 2) with oven drying methods. For respectively, indicating Method I is suitable only for
feeds which ≤ 6.6% urea, volatilization of urea during nonurea feeds or feeds containing less than 5% urea.
drying for Method B, Method J, Method I and Method The correlation coefficients for feed with urea and feed
L was small enough to keep water recoveries at or below without urea for Karl Fischer vs. the Method L were
105%. However, significant urea losses were observed 0.76 and 0.98, and the slopes were 1.09 and 1.19, respec-
with Method A, even at less than a concentration of tively, indicating no general improvement over
1% urea. Method J.

Downloaded from jas.fass.org by on October 6, 2010.

 Measuring moisture content of feeds 3263
tein nitrogen was observed for any of the methods eval-
uated (Table 5). The mean biases reported by Shreve
et al. tend to mask the large range in minimum to
maximum bias and % recovery. Even Method B, which
was not statistically different from Karl Fischer had a
range in bias from -0.69 to 1.73 (range in recovery from
112% to 73%).
The estimation of moisture in feeds with oven meth-
ods is frustrating. Decomposition of urea from urea
feeds is observed with all oven methods and becomes
very problematic at a concentration of 10% urea.
Method B (104°C, 3 h) method most closely approxi-
mated values obtained by Karl Fischer in urea con-
taining feeds. All other oven methods overestimated
moisture in urea feeds. The nonurea feeds pose a num-
ber of problems from decomposition of milk replacer at
135°C, to incomplete removal of water in other prod-
ucts. Overall, Method A (135°C, 2 h) and Method L
110°C, 3 h methods most closely approximated Karl
Fischer values. The bound water in corn, soybeans, and
barley was difficult to remove, even at 135°C. Given
the problems associated with the 135°C temperature,
we would recommend using a temperature not to exceed
110°C for any material. The NFTA method (Method
B, 105°C for 3 h) method could potentially be used to
estimate moisture in feeds but should be validated
Figure 2. Recovery of water as a function of urea con-
against Karl Fischer for each type of feed matrix.
centration (for oven methods based on Karl Fischer).
Comparison of Oven Methods to Karl Fischer Method
for Steam-Flaked Corn and Corn Products
Comparisons made by Shreve et al. (2000), found
mean biases of 0.24, 1.10, 0.92, and 0.51 for the Method Previous research by investigators has shown that
B, Method N, Method J, and Method O, respectively, steam flaking of corn and grain sorghum improves
as compared to Karl Fischer (See Table 5). The Karl starch utilization and feed conversion by feedlot cattle,
Fischer and Method B produced comparable moisture and that moisture level and temperature are critical
values (α = 0.01). All vacuum oven methods tested re- factors in controlling the quality of flakes produced
sulted in significantly lower moisture content compared (Richardson, 1996a; Richardson, 1996b). Moisture in
to the Karl Fischer moisture. No relationship of nonpro- whole shelled corn and in the same source of corn after

Table 5. Comparison of loss-on-drying methods (LOD) and Karl Fischer method in urea supplements
Oven method

B N J O
104°C/3 h 60°C vacuum/20 h 95°C vacuum/5 h 70°C vacuum/20 h

Difference and % recoverya,b

Item n Biasa % recoveryb Biasa % recoveryb Biasa % recoveryb Biasa % recoveryb

Feed grade ureac 1 −0.50 398 0 97 −0.49 396 0.17 0

Mean bias of one feed grade urea
and 10 supplements 11 0.24 95 1.10d 81 0.92d 84 0.51d 91
Minimum bias and % recovery .07 73 0 68 0.27 71 −.08 76
Maximum bias and % recovery 1.73 112 2.05 97 1.87 95 1.57 101
SD 0.75 — 0.63 — 0.72 — 0.55 —
tdiffe 1.05 — 5.84 — 4.22 — 3.10 —
a
Bias is Karl Fischer moisture minus each respective LOD method.
b
Recovery as a percent of Karl Fischer.
c
Moisture content of feed grade urea by Karl Fischer method was 0.17%.
d
Bias with superscript indicates the mean differs from the Karl Fischer Mean, df = 10, t at α = 0.01 of 2.76; t-test significant if tdiff > t at
α = 0.01.
e
Student’s t calculated on the difference between pairs of data (paired t-test).

Downloaded from jas.fass.org by on October 6, 2010.

3264 Thiex and Richardson

Table 6. Moisture content of corn and corn products determined

by Karl Fischer and oven dryinga
Karl Fischer Oven method Bias % Recovery

Feedstuff n KF1 M (110°C/15 h)

Moisutre content, %
Corn
Whole, shelled 3 14.75 13.64 −1.11 92.5
Steam-flaked, (warm up) 3 25.24 25.86 .62 102
Steam-flakedb 3 21.74 23.43 1.69 108
Steam-flakedb 3 23.25 23.97 .72 103
Steam-flakedc 3 23.57 22.19 −1.38 94
Steam-flakedc 3 21.26 20.79 −.47 98
Steam-flaked, mean 23.01 23.25 0.23
a
Unpublished data by Richardson (2002). Oven drying, Texas Tech Univ., Karl Fischer, South Dakota
State Univ., no statistical treatment.
b
The steam flaker was a small scale, pilot unit, with a round steam cabinet.
c
Steam flaker was a commercial model with a rectangular steam cabinet.

steam flaking with two different flaking units is pre- were 0.203 and 0.980 for Karl Fischer, 0.504 and 0.958
sented in Table 6. The average bias of percent moisture for Method A, 0.356 and 0.968 for Method B, and 0.412
determined by Method M and by Karl Fischer for steam- and 0.962 for the Method I respectively. These results
flaked corn was small (0.23). However, a greater bias indicate that Karl Fischer is the method of choice for
for whole shelled corn was observed (−1.11) than for NIR calibrations.
steam-flaked samples. These data indicate that oven Results of NIR calibrations for feed materials and for
drying slightly overestimated moisture in steam-flaked all materials combined by Karl Fischer are reported in
corn, and underestimated moisture in whole shelled Table 8. Partial least squares calibration equations
corn by 8.5%, as compared to the Karl Fischer method. were generated using a 1, 4, 4, 1 math treatment. The
The difference between the two methods was greater SEC and R2 were 0.358 and 0.980 on feed materials
for whole shelled corn and the use of the Karl Fischer alone for Karl Fischer, and 0.298 and 0.982 on all feed
method offers potential for improving accuracy of mois- and forage materials combined, respectively.
ture determination. Preliminary NIR calibrations were made for two rea-
Results of our evaluation of the preanalyzed samples sons: 1) to develop an idea of the accuracy of the refer-
submitted by the Corn Refiners Association are shown ence method to measure water (good calibrations can-
in Table 7 and indicate very good agreement between not be obtained with a poor reference method), and 2)
Karl Fischer method KF1 and method P. The Associa- to determine the feasibility of using Karl Fischer as a
tion concluded that the Karl Fischer method could be potential reference method for NIR calibrations for the
adopted as a primary method for the standardization determination of moisture in a diverse set of animal
of secondary loss-on-drying methods. feed materials.
NIR calibrations for moisture in forages were made
Assessing the Potential for NIR with no effort to optimize the calibrations, or to select
Calibrations for Moisture in Feeds samples for calibration based on spectra. Simply, the
samples used for the experiments were used to deter-
Results of NIR calibrations for forage materials by mine the feasibility and relative accuracy of NIR cali-
four methods are reported in Table 8. The standard brations for moisture (dry matter) based on Karl Fis-
error of calibration (SEC) and correlation coefficient (r) cher and the three oven methods. Best calibrations were

Table 7. Corn Refiners Association comparison of loss-on-drying methods

to Karl Fischer method for corn productsa
Method: Method KF1 (Karl Fischer) Method P (100°C vacuum/4 h)

Laboratory: SD State University Corn products Roquette

Feedstuff Mean Recheck Recheck

Corn gluten meal, 3/99 12.4 (0.1) 12.63 (0.58) 13.1 13.1
Corn gluten meal, 4/99 9.38 (0.17) 9.31 (0.09) 9.31 9.4
Corn gluten feed pellets 8.06 (0.14) — 8.1
Corn gluten meal 8.33 (0.09) — 8.4
a
Values given as moisture content, % (SD).

Downloaded from jas.fass.org by on October 6, 2010.

 Measuring moisture content of feeds 3265
Table 8. Near Infrared Reflectance calibrations for moisture in feeds,
forages, and all samples based on Karl Fischera
No. of No. of Math
Item samplesb PLS termsc treatmentd SECc R2 f SECVg 1-VRh

Feeds KF1 14 5 1,4,4,1 0.358 0.980 0.988 0.850

Forages
KF1 (Karl Fischer) 32 5 1,4,4,1 0.203 0.980 0.306 0.959
A (135°C/2 h) 32 5 1,4,4,1 0.504 0.958 0.870 0.878
B (104°C/2 h) 31 5 1,4,4,1 0.356 0.968 0.533 0.934
I (104°C/2 h) 31 5 1,4,4,1 0.412 0.962 0.602 0.925
All samples KF1 (Karl Fischer) 45 6 1,4,4,1 0.298 0.982 0.403 0.968
a
Adapted from Thiex and Van Erem (1999).
b
Number of samples after removal of outliers.
c
Number of terms in the Partial Least Squares calibration equation.
d
The first value indicates the derivative (2 = second derivative), the second value indicates the gap over
which the derivative is calculated, the third value is the number representing the smoothing of points, and
the fourth value indicates a second smooth (1 = no second smooth).
e
Standard error of calibration.
f
Correlation coefficient.
g
Standard error of cross validation.
h
One minus the variance ratio (explained variance divided by the total variance).

obtained using Karl Fischer because calibrations could based on Karl Fischer as the primary method should
be made directly for water (best calibration data), while replace oven methods as a secondary method for routine
calibrations for the oven methods would be based on moisture estimates.
water and various other volatile components of the for-
age materials. Literature Cited
NIR calibrations based on Karl Fischer were also
Analytical Methods of the Member Companies of The Corn Refiners
attempted for feeds even though the population was
Association, Inc. 1999. 6th ed. CRA, Washington, DC.
small (14 samples). Considering the diversity of the AOAC International. 2000. Official Methods of Analysis. 17th ed.
samples, different grinding procedures, and small popu- AOAC Int., Gaithersburg, MD.
lation, good calibrations were obtained indicating excel- Bernetti, R., S. J. Kochan, J. J. Pienkowski. 1996. Determination of
lent feasibility for robust calibrations for moisture in moisture in corn by the Karl Fischer Method, Collaborative
Study. Poster session, AOAC, Washington, DC. Method also
animal feed. The 33 forage samples and 14 feed samples
available as CRA Analytical Method A-13 Moisture (Karl Fis-
were combined to determine the feasibility of a mixed cher) in Corn; revised Nov. 17, 1992.
calibration. Calibrations were acceptable for combined Jones, F. E. 1985. Mass basis Karl Fischer titration equation for
feed and forage samples, indicating one calibration for moisture determination. J. AOCS 62:1605.
forages and feeds is feasible with additional work. For Kellems, R. O. and D. C. Church. 2002. Pages 39–54. Feedstuffs. In:
Livestock feeds and feeding 5th ed. Prentice Hall, Upper Saddle
laboratories with NIR capabilities, NIR calibrations for River, New Jersey.
moisture based on Karl Fischer appear to be a vast Horwitz, W. A., R. Albert, M. J. Deutsch, and J. N. Thompson. 1990.
improvement over oven methods. Precision parameters of methods of analysis required for nutri-
tion labeling. Part I. Major nutrients. J. AOAC 73:661–680.
Implications Hunt, W. H. and S. W. Pixton. 1974. Pages 1–55 in Storage of Cereal
Grains and Their Products. American Association of Cereal
Chemists, St. Paul, MN.
Improving moisture measurements is necessary to Mo, M., and T. Tjornhom. 1978. Losses of carbon-containing sub-
improve dry matter calculations of other nutrients, stances during dry matter determination by oven drying. Acta
thereby improving diet formulations. We recommend Agric. Scand. 28:196–202.
NFTA. Moisture Task Force reports to NFTA Board of Directors,
that users of oven methods be aware of the limitations
February 2001. Natl. Forage Testing Assoc., Omaha, NE.
and error associated with the methods, and never apply Richardson, C. R. 1996a. Feedmill operation, feed handling, pro-
methods to materials for which they have not been vali- cessing and good manufacturing practices. Pages 129–144 in
dated. Results obtained using oven methods should not Cattle Feeding: A Guide to Management. Trafton Printing, Inc.,
be termed “moisture.” They should be termed “loss on Amarillo, TX.
Richardson, C. R. 1996b. Quality control in feed production. Pages
drying,” and the drying conditions should be part of the 1–5 in Proc. 1996 Mid-South Nutrition Conf., Dallas, TX.
term, for example, “loss on drying, 104°C/2 h, 5.0%.” Robertson, J. A., and W. R. Windham. 1983. Automatic Karl Fischer
Drying at 135°C should be eliminated or restricted to titration of moisture in sunflower seed. J. Am. Oil Chem. Soc.
very few materials for which it has been validated. The 60:1773–1777.
Karl Fischer method should be routine in feed labora- Shreve, B. R., R. Royle, and N. Thiex. 2000. Comparisons of loss-on
drying methods for dry matter determination on feed and forage
tories, both as a final method and as a primary method samples. Abstracts 2000 Midwest Section AOAC Int. and 2000
to evaluate loss on drying methods before they are prac- National Forage Testing Association Workshop, Abstract #5.
ticed. Near Infrared Reflectance calibrations for feeds Sioux Falls, SD.

Downloaded from jas.fass.org by on October 6, 2010.

3266 Thiex and Richardson
Thiex, N. 2002. Committee on feeds, fertilizers, and related agricul- Van Erem, T., N. Thiex, J. Pohmer, W. M. Poffenbarger, V. Smith,
tural topics. Feeds. J. AOAC 85:270–273. and E. Patel. 1998. Determination of water in forages and animal
Thiex, N., and T. Van Erem. 1999. Comparisons of Karl Fischer feeds by Karl Fischer titration. J. AOAC 81:25–32.
method with oven methods for determination of water in forages Windham, W. R., and F. E. Barton II. 1991. Moisture analysis in
and animal feeds. J. AOAC 82:799–808. forage by near- infrared reflectance spectroscopy: Collaborative
Thiex, N., and T. Van Erem. 2002. Determination of water (moisture)
study of calibration methodology. J. AOAC 74:324–331.
and dry matter in animal feed, grain and forage (plant tissue)
by Karl Fischer titration: Collaborative Study. J. AOAC Windham, W. R., J. A. Robertson, and R. G. Leffler. 1987. A compari-
85:318–327. son of methods for moisture determination of forages for new
Undersander, D., D. R. Martens, and N. Thiex. 1993. Forage Analysis infrared reflectance spectroscopy calibration and validation.
Procedures. Natl. Forage Testing Assoc., Omaha, NE. Crop. Sci. 27:777–783.

Downloaded from jas.fass.org by on October 6, 2010.

Citations This article has been cited by 1 HighWire-hosted articles:
http://jas.fass.org/cgi/content/full/81/12/3255#otherarticles

Downloaded from jas.fass.org by on October 6, 2010.