You are on page 1of 7

Corn Moisture Meter-To-Oven Comparisons

M. R. Paulsen, L. D. Hill, B. L. Dixon


MEMBER
ASAE

ABSTRACT sample variabihty, with the remaining 20% attributed to


FTER revised calibrations for electronic moisture meters moisture meter and oven variability.
A were placed into effect on August 1, 1981 for Illinois and
Iowa, the accuracy of the calibrations were tested during the OBJECTIVES
Fall 1981 harvest using natually wet corn samples. At the The objectives of this study were:
University of IlHnois the moisture content of 690 samples was 1. To determine the bias of corn moisture meters using the
measured three times in each of six commonly used commercial 1981 IlHnois-Iowa caHbrations compared to the USDA air-oven
moisture meters, additional data from the IlHnois Department reference method.
of Agriculture is presented for 176 samples tested in 10 2. To determine 95-percent confidence intervals for allowable
commercial moisture meters. Moisture contents of the samples tolerances between moisture meter and air-oven moisture
ranged form 11 to 36%. measurements.
Moisture meter bias was significantly affected by hand versus 3. To test for regional differences in moisture meter
combine shelling for some meters. Ninty-five percent confidence performance.
hmits for meter tolerances typically ranged from ±0.8
percentage points at low moistures to ± 3.2 percentage points PROCEDURE
at high moistures. Motomco moisture meter accuracy at
Samples Collected
moistures above 22% was improved by the calibration changes.
During Fall 1981, 690 corn samples were tested in six
commercial moisture meters. The samples were obtained from
INTRODUCTION
State Departments of Agriculture in Arkansas, Colorado,
This study was conducted as a continuation of the electronic Delaware, Georgia, Illinois, Kentucky, Maryland, Mississippi,
moisture meter-to-oven comparisons performed in Fall 1980. Nebraska, South Carohna, Tennessee, and Wisconsin; from
As a result of the Fall 1980 study, similar results from Hurburgh FGIS at Grandview, Missouri; IHinois seedcorn plots; and
et al., (1981) and moisture meter manufacturers' tests, new commercial grain elevators in IlHnois, Indiana, and Ohio (part
cahbrations for high moisture content corn were placed into of Ohio elevator receipts were from Michigan). Samples ranged
effect for all commercial meters used in IHinois and Iowa after from 11 to 36% moisture*. High-moisture samples from State
August 1, 1981. On September 1, 1981, the revised calibration Departments of Agriculture were placed into sealed jars packed
charts for the Motomco moisture meter became nationwide in ice and air-shipped in polystyrene coolers. They were usually
when the Federal Grain Inspection Service (FGIS) of USDA tested immediately after warming or on the following day after
provided tentative approval for all Federally and State licensed receipt.
grain grading inspectors to use C-12 and C-13 charts instead
of the former C-2-D and C-3-B high moisture charts. Meters Tested
Hurburgh et al. (1981) conducted moisture meter-to-oven The samples were tested in six commonly used commercial
comparisons on 879 samples of 1979 and 1980 corn. They moisture meters, Motomco 919t, Dickey-john Gac-II, Burrows
reported that all meters tested had significant cahbration 700, and Steinlite models SS-250 DM, and RCT as shown in
biases ranging from 1.5 points above to 3.5 points below the Fig. 1.
oven moisture determination. They also found that about 80%
of the variance between meter-to-oven comparisons was due to Laboratory Procedures
The laboratory procedure consisted of uniformly mixing the
sample by dividing with a Boerner divider and re-mixing three
Article was submitted for publication in January, 1984; reviewed and times. To eHminate foreign material (FM) as a potential factor
approved for publication by the Electric Power and Processing Div. affecting moisture meter readings, the FM was removed by
of ASAE in June, 1984. Presented as ASAE Paper No. 82-3548.
Authors are: M. R. PAULSEN, Associate Professor, Agricultural mechanically sieving with a 4.76 mm (12/64 in.) round-hole
Engineering Dept.; L. D. HILL, Professor, and B. L. DIXON, sieve. Next, test weight was determined and the sample was
Associate Professor, Agricultural Economics Dept., University of split into fourths with the Boerner divider. One one-fourth
Illinois, Urbana. portion was tested with three drops in each of three meters. Due
Acknowledgments: This study was a part of Project No. 39062 of
the Agricultural Experiment Station, College of Agriculture, University to differing sample sizes required, the following order was
of Illinois at Urbana-Champaign. It was supported in part by grants maintained: Drop 1 GAC-II, Burrows 700, and Motomco 919
from Illinois Grain and Feed Association, and the Illinois Department — Drops 2 and 3 Motomco 919, Burrows 700, Burrows 700,
of Agricultural Division of Agricultural Industry Regulation. Equipment
and/or technical assistance was provided by Dickey-john Corp.; Fred
Stein Laboratories, Inc.; Motomco, Inc.; Seedburo Equipment Co.; *A11 moisture contents used in this study are expressed as a percent
Shore Sales Co. Inc.; DeKalb Ag Research, Inc,; the Federal Grain wet basis.
Inspection Service, USDA; the Arkansas, Colorado, Delaware, Georgia, tTrade names are used in this publication solely for the purpose of
IlHnois, Kentucky, Maryland, Mississippi, Nebraska, South Carolina, providing specific information. Mention of a trade name, proprietary
Tennessee, and Wisconsin Departments of Agriculture; and Iowa State product or specific equipment does not constitute a guarantee or
University. Appreciation is expressed to members of the Illinois-Iowa warranty by the University of Illinois or the Illinois Department of
Task Force, numerous grain elevators that provided samples, and all Agriculture and does not imply approval of the named product to the
University of Illinois personnel who assisted with the research. exclusion of other products that may be suitable.

1984—TRANSACTIONS of the ASAE © 1984 American Society of Agricultural Engineers 0001-2351/84/2706-1917$02.00 1917
TABLE 2. ESTIMATED ILLINOIS-IOWA CORN MOISTURE
METER CALIBRATION CHANGES FROM CALIBRATIONS
IN EFFECT DURING FALL 1980
Corn
moisture, %
Type of
Meter change 22 30

Motomco 919 chart +0.24 +2.50


Burrows 700 Slope, intercept +0.60 +1.00
Dickey-john GAC-11 K-values 0.00 + 2.00
Steinlite
SS-250 Test weight +0.19 +0.47
DM Test weight +0.14 +0.36
AUT None 0.00 0.00
Fig. 1—Moisture meters tested in 1981 were (L-R) the Dickey-john DL Chart +0.08 +0.93
RCT Chart 0.00 +1.04
GAC-II, Burrows 700, Motomco 919, and Steinlite Models SS-250, DM RC
and RCT. Chart 0.00 +0.93
G-400 Chart 0.00 +0.47
Motomco 919, GAC-II, and Gac-II. Another one-fourth portion
was tested with three drops in the three SteinUte meters. The for Fall 1981 from the previous 1980 calibrations are shown
following order was maintained: Drop 1 SS-250, DM, and RCT; in Table 2, Wollin et al., (1982). It was estimated that the Fall
Drop 2 RCT, DM, and SS-250; and Drop 3 SS-250, DM, and 1981 calibrations increased the meter readings over the 1980
RCT. Another one-fourth portion was used for oven moisture calibrations by as much as 2.5 percentage points at 30 percent
determinations. The final one-fourth portion was kept as a '*file moisture. At moistures below 22%, calibration changes were
sample" in refrigerated storage for about a week for future very small if not negligible. In fact, for the Motomco meter,
testing if necessary. used by FGIS inspectors, the low moisture chart (8.0 - 21.09%)
was unchanged in Fall 1981.
Meter Calibrations and Correction Factors
The calibrations and correction factors for test weight and Oven Moisture Determinations
grain temperature in effect in Illinois and Iowa beginning in The USDA air-oven method of determining moisture was
Fall 1981 are shown in Table 1. The correction factors were used on three sub-samples taken from the one-fourth sample
added or subtracted from initial moisture values either portion. The USDA procedure prescribes that the sub-samples
automatically by the meters or by a computer program when be dried at 103 ± 1 °C for 72 h in a forced-draft oven (USDA,
the data were analyzed. The estimated change in meter readings 1971; ASAE, 1983). It states that 15-g samples be used for

TABLE 1. SUMMARY OF MOISTURE METER CALIBRATIONS, TEMPERATURE AND TEST WEIGHT CORRECTION FACTORS USED
FOR SHELLED CORN IN ILLINOIS DURING FALL 1981
Test weight correction
(added to temperatu re corrected
Calibration moisture for each one lbA)u
Temperature correction different than £
Applicable 6 Ib/bu
Meter moisture Sample (added to moisture Moisture Less than More than
model range, % size, g Method per°F) range, % 56 Ib/bu 56 Ib/bu

Motomco 919 8.0 - 21.09 250 Chart C-l-C above 77 °F +0.052 None
below 77 °F -0.052
21.09- 29.71 250 Chart C-12 above 77 °F +0.052 None
below 77 °F -0.052
29.71- 40.21 150 Chart C-13 above 77 °F +0.024 None
below 77 °F -0.024
Burrows 700 10.0 - 35.0 250 Slope = 97.5 Automatic None
Intercept =95.7
Dickey-john 7.0 - 22.0 — Kl - K9 values Automatic Automatic
GAC-II 22.01 - 40.2 Kl - K9 values
Steinlite 10.0 - 25.0 250 No. 4 module Automatic 10.0 - 21.0 +0.10 -0.05
SS-250 25.01 - 35.0 250 No. 4 HM module 21.01 - 35.0 +0.15 -0.10
Steinlite 10.0 - 25.0 250 No. 3 module Automatic 10.0 • 20.0 +0.15 -0.15
DM 25.01 - 35.0 250 HM module or HM chart 20.01 • 35.0 +0.07 -0.07
Steinlite 10.0 - 25.0 250 No. 4 plate Automatic 10.0 • 20.0 +0.12 -0.12
Automatic 25.01 - 35.0 250 No. 4 plate and HM chart 20.01 • 35.0 0 0
Steinlite 7.0 - 35.9 250 No. 3 chart above 80 °F +0.05 10.0 • 20.0 +0.10 -0.10
DL below 80 °F -0.05 20.01 - 35.0 0 0
Steinlite 8.4 -35.8 250 No. 3 HM chart above 80 °F +0.05 8.5 • 22.0 +0.08 -0.06
RCT below 80 °F -0.05 22.01 - 35.8 +0.16 0
Steinlite 8.5 -35.3 250 No. 3 HM chart above 80 °F +0.05 8.5 - 22.0 +0.08 -0.06
RC below 80 °F -0.05 22.01 • 35.3 +0.16 0
Steinlite 7.1 -35.5 100 No. 3 HM chart above 80 °F +0.05 7.1 • 22.0 +0.10 -0.10
G-400 selector button below 80 °F -0.05 22.01 • 35.5 0 0

1918 TRANSACTIONS of the ASAE—1984


TABLE 3. OVEN MOISTURE COMPARISONS FOR 1981
CORN DRIED AT 103°C FOR 72 HOURS
Sample Arkansas Dept Dickey-john University of
no. of Agriculture Laboratory Illinois
Sample size
15g 15g 15g/100g
1 30.36 30.83 30.71
2 20.98 20.75 20.99
3 19.08 19.57 19.79
4 24.98 24.63 24.74
5 23.23 23.39 23.30
12 20.93 21.07 21.36
13 21.48 21.56 21.76
14 16.65 16.75 16.74
15 21.54 21.65 21.67
16 23.83 24.17 24.19
Mean 22.31 22.44 22.53

Fig. 2—Forced-convection ovens used included two Blue-M Model


POM7-256C units. Statistically analyzed using two-tailed paired t-tests to test the
hypothesis (H^: fi^ = /x^) that there was no difference in meter
moisture less than or equal to 25% and 100-g samples be used performance between the mean for the hand-shelled and the
for moistures above 25%. The Motomco moisture was used to combine shelled samples.
decide whether 15-g or 100-g oven samples were needed. After
samples were removed from the oven, their respective lids were Statistical Procedures
placed on the pans, the samples were allowed to cool to room It was desired to construct 95-percent confidence intervals
temperature in dessicators using activated alumina as a and obtain prediction equations for meter-minus-oven values
dessicant. Samples were weighed on an analytical balance to (meter bias denoted as Y below) versus oven moisture content.
the nearest 0.1 mg. If the three sub-samples varied by more than Since the meter bias appeared to increase for increasing oven
0.4 moisture percentage points, they were rerun. All samples moistures, Barlett's test for homogeneity of variances was
were dried in one of two Blue-M Model POM 7-256C ovens, performed (Snedecor and cochran, 1971). Barlett's test of the
which had a solid-state proportional heating controller that null hypothesis of homoskedasticity was rejected at the 1-percent
maintained uniform temperature in the oven. Fig. 2. level of error. Thus, the procedure described in Paulsen et al.,
Oven standardizations were performed on samples exchanged (1983) was used to divide meter bias values into 4-percentage
with the Illinois Department of Agriculture (ILDA); Dickey- point moisture intervals. Generalized least squares (GLS)
john laboratories; FGIS at Grandview, MO; Iowa State estimates of the regression coefficients, a and b, for the Unear
University; and several State Departments of Agriculture. regression models were obtained after dividing all variables by
the standard deviation of Y for each 4-point moisture interval,
Illinois Department of Agriculture Procedures SAS (1982). Next, the variance-covariance matrix of the GLS
The Illinois Department of Agriculture conducted a moisture estimates plus the error term variance were used to obtain a
meter testing program with 176 corn samples from Carroll, forecast confidence interval as described in Paulsen et al.,
Christian, Clay, Coles, Edwards, Franklin, Gallatin, Logan, (1983).
Macoupin, Madison, Marshall, Menard, Monroe, Montgomery,
Saline, Sangamon, St. Clair, Stephenson, Tazewell, Wayne, RESULTS AND DISCUSSION
White, and Whiteside counties in Illinois (Schwarberg et al.,
Oven Standardization
1982). The corn was hand-shelled, cleaned with a 4.76 mm
Oven standardization test results for 10 corn samples air-
(12/64 in.) round-hole sieve, and mixed four times with a
shipped from Arkansas Department of Agriculture to the
Boerner divider. The following ten moisture meters were tested:
University of lUinois and to the Dickey-john laboratories in
Motomco 919, Dickey-john GAC-II, Burrows 700, and Steinlite
Auburn, IL are presented in Table 3. The means for three oven
Models SS-250, Automatic, DM, DL, RCT, RC and G-400.
determinations for 10 samples averaged within 0.22 and 0.09
Three sample drops were performed on each meter. Drop 2 was
percentage points of the Arkansas and Dickey-john laboratories,
performed in reverse order than drops 1 and 3. Three 15-g
respectively.
moisture determinations were performed on each sample using
University of Illinois oven tests on 6 samples from the South
the prescribed USD A air-oven reference method. Data for meter
CaroUna Department of Agriculture averaged within 0.13
and oven moistures, test weight, and sample temperatures were
percentage points of each other. Table 4.
provided to the Agricultural Engineering Department at the
University of Illinois (UI) for statistical computation and
analysis. TABLE 4. OVEN MOISTURE COMPARISONS
FOR 1981 CORN DRIED
AT 103° C FOR 72 HOURS
Hand-Shelling Versus Combine-Shelling
Sample S. Carolina Dept. University of
To determine if moisture meter performance might differ no. of Agriculture Illinois
between hand-shelled and combine-shelled samples, 20
commercial corn hybrids were both hand-picked and combine 171 13.38 13.23
172 12.18 12.10
harvested from 20 plots at Salem, IL on the same day. Ear- 173 22.77 22.53
corn and combined samples from the same field location were 174 15.24 15.20
tightly enclosed in plastic bags to prevent moisture loss prior 175 21.93 21.96
176 17.62 17.35
to moisture meter testing. Moistures ranged from 21 to 29%.
Mean 17.19 17.06
Meter-minus-oven values for each pair of samples were

1984—TRANSACTIONS of the ASAE 1919


TABLE 5. OVEN MOISTURE COMPARISONS FOR 1981 CORN
DRIED AT 103° C FOR 72 HOURS
Sample Illinois Dept. University Dickey-john Iowa State FGIS
no. of Agriculture of Illinois Laboratory University
158 30.16 30.31 29.87 31.20 30.54
159 24.54 24.86 24.64 25.10 25.20
160 26.57 27.31 26.59 27.40 27.02
Mean 27.09 27.49 27.03 27.90 27.59
478 25.03 24.89
479 25.84 25.79
480 26.06 26.00
481 25.51 25.53
482 26.83 26.79
483 25.99 26.00
Mean 25.88 25.83
618 22.32 22.53
619 22.36 22.37
620 23.70 23.69
621 23.68 23.92
Mean 23.02 23.13

Samples for oven tests were exchanged over three different On an additional 14 samples between 12 to 15% moisture.
time periods with ILDA, Table 5. Tests averaged within 0.40, University of Illinois ovens deviated by +0.28 to - 0 . 3 4
0,05, and 0.11 percentage points of each other during the three percentage points of the FGIS ovens at Grandview, MO. The
periods. During the first period, UI oven determinations UI and FGIS ovens averaged within 0.03 percentage points of
averaged within 0.46, 0.41 and 0.10 percentage points of the each other.
Dickey-john, Iowa State University and FGIS laboratories,
respectively. Meter Bias
Data are presented in Figs. 3 through 8 for samples from all
MOTOMCO locations for the meters tested at the University of Illinois. Data

5 B700

18 20 22 24 26 28 38
OVEN MOISTURE, X
28 22 24 26 28
Fig. 3—Moisture meter bias versus oven moisture for University of OVEN MOISTURE, X
Illinois 1981 tests for the Motomco 919. (Symbol "x" denotes the
95-percent confidence limits about the predicted line which is denoted Fig. 5—Moisture meter bias versus oven moisture for University of
by ' ' o " symbols.) Illinois 1981 tests for the Burrows 700.

5 r GACII SS250

4h
3

12 M 16 18 29 22 24 26 28 30 32 34 36 12 14 16 18 20 22 24 26 28 30 32 34 36
OVEN MOISTURE, X OVEN MOISTURE, X

Fig. 4- -Moisture meter bias versus oven moisture for University of Fig. 6—Moisture meter bias versus oven moisture for University of
Illinois 1981 tests for the Dickey-john GAC-II. Illinois 1981 tests for the Steinlite SS-250.

1920 TRANSACTIONS of the ASAE—1984


5 DM TABLE 7. GENERALIZED LEAST SQUARES ESTIMATES OF
REGRESSION COEFFICIENTS a AND b AND STANDARD
4 ERRORS FOR 1981 CORN TESTED BY
ILLINOIS DEPARTMENT OF AGRICULTURE
3

< Meter model a Std error b Std error


, 2 X.

u X Motomco 919 1.15 0.40 -0.068 0.016


D 1
X .; ..»••. . ; • X . Dickey-john
o
2 0 ' . •>•
GAC-II 1.80 0.30 -0.070 0.012
'.•.••f;';'i'i--.'ii:'..-i-." .•iv'- •. '• - Burrows 700 5.09 0.38 -0.202 0.015
.;?/-,r.- •?. '•
' • x •><
Steinlite
SS-250 2.83 0.29 -0.118 0.011
^'
^ • ' •.•
c-2 DM 2.87 0.26 -0.124 0.011
-3
X Automatic 3.64 0.29 -0.153 0.012
RCT 3.28 0.32 -0.146 0.013
-4 - RC 4.34 0.27 -0.195 0.011
DL 3.91 0.26 -0.145 0.011
-5
12 M 16 18 20 22 24 26 28 38 32 34 36
G-400 3.25 0.45 -0.099 0.018
OVEN MOISTURE, X

Fig. 7 - -Moisture meter bias versus oven moisture for University of are presented in Tables 6 and 7 for the model, meter-minus-
Illinois 1981 tests for the Steinlite DM.
oven moisture = a + b (oven moisture). (A regression
coefficient is statistically significant at approximately the 95 ^o
RCT
level if the absolute value of the ratio of the estimated coefficient
to its standard error is greater than 1.96).
For the UI data nearly all of the meters have positive slope
coefficients and negative intercept coefficients in 1981, Table
6. A positive slope coefficient means that as oven moisture
increases the meter reading bias moves in a positive direction.
However, for the ILDA data all of the meters have negative
slope and positive intercept coefficients. Table 7. A possible
explanation of this apparent discrepancy in meter readings is
that all of the ILDA samples were hand-shelled and from
IlHnois. If we look only at UI samples that were hand-shelled
and from Illinois, the slope coefficients are negative and the
intercept coefficients are positve as shown in Table 6. Further
12 14 16 18 28 22 24 26 28 30 32 34 36 examination of Table 6 reveals that combine-shelled corn from
OVEN MOISTURE, X IlHnois, Indiana, Ohio, and from all locations typically have
positive slope coefficients. This evidence suggests that electronic
Fig. 8—Moisture meter bias versus oven moisture for University of
Illinois 1981 tests for the Steinlite RCT. moisture meters tend to read drier as oven moisture increases
if the samples are hand-shelled instead of combine shelled.
are presented in Paulsen et al. (1982) for the meters tested by A further test of this hypothesis on the 20 samples from
ILDA. The figures present individual data points for meter- Salem, IL that were both hand and combine-shelled, revealed
minus-oven values as a function of oven moisture content. Each that the Dickey-john GAC-II and the Steinlite RCT read
data point is the mean of three drops in a meter and three oven significantly lower at the a = 0.05 level on hand-shelled corn
tests. GLS prediction Hues are plotted for each meter. The than on combine-shelled corn of the same variety (Table 8). The
regression coefficients and standard errors for the coefficients Burrows 700 and the Motomco 919 read significantly lower at

TABLE 6. GENERALIZED LEAST SQUARES ESTIMATES OF REGRESSION COEFFICIENTS, a AND b, AND STANDARDS ERRORS FOR 1981 CORN
TESTED AT THE UNIVERSITY OF ILLINOIS

All locations Illinois Illinois Indiana Ohio All locations


hand- hand- combine combine combine combine
shelled shelled shelled shelled shelled shelled All samples

Number of samples 107 45 211 108 52 416 636

Meter model

a s.e. a s.e. a s.e. a s.e. a s.e. a s.e. a s.e.


Motomco 9 1 9 -0.56 0.12 0.73 0.45 -0.08 0.16 -1.55 0.28 -2.68 0.18 -1.72 0.12 -1.22 0.12
Dickey-jo h n
GAC-II 0.34 0.10 1.18 0.47 -1.08 0.19 -2.71 0.26 -3.94 0.27 -1.99 0.13 -1.55 0.13
Burrows 700 -0.07 0.29 1.27 0.37 -1.22 0,25 -4.33 0.26 -3.09 0.25 -2.36 0.14 -2.02 0.14
Steinlite
SS-250 1.71 0.37 1.30 0.42 -0.06 0.19 -1.89 0.21 -1.75 0.23 -1.49 0.05 -1.46 0.04
DM 1.93 0.31 0.58 0.61 -0.06 0.14 -0.50 0.18 -0.95 0.20 -0.59 0.11 -0.49 0.11
RCT 1.78 0.38 1.24 0.38 -0.89 0.18 -2.59 0.23 -3.15 0.28 -2.05 0.13 -1.58 0.13

b s.e. b s.e. b s.e. b s.e. b s.e. b s.e. b s.e.


Motomco 9 1 9 0.028 0.008 -0.035 0.022 0.009 0.009 0.072 0.015 0.126 0.008 0.085 0.006 0.058 0.006
Dickey-jo h n
GAC-II -0.006 0.007 -0.045 0.020 0.063 0.009 0.150 0.014 0.195 0.012 0.108 0.007 0.085 0.006
Burrows 700 0.020 0.014 -0.054 0.017 0.074 0.011 0.221 0.013 0.143 0.011 0.121 0.007 0.106 0.007
Steinlite
SS-250 -0.061 0.016 -0.051 0.018 0.015 0.009 0.098 0.011 0.079 0.010 0.075 0.003 0.073 0.003
DM -0.088 0.013 -0.032 0.025 -0.008 0.007 -0.001 0.009 0.014 0.009 0.008 0.005 0.005 0.005
RCT -0.073 0.017 -0.059 0.018 0.049 0.008 0.127 0.012 0.145 0.012 0.010 0.006 0.077 0.006

1984—TRANSACTIONS of the ASAE 1921


TABLE 8. EFFECT OF HAND-VERSUS-COMBINE SHELLING ON TABLE 9. TEST WEIGHT DISTRIBUTION (kg/m^)
METER-MINUS-OVEN VALUES FOR 20 COMMERCIAL BY FOUR-POINT MOISTURE INTERVALS
CORN HYBRIDS FROM SALEM, IL IN 1981
Hand- Combine- Calculated Oven moistiire range, %
Meter model shelled shelled t-value
8-12 12-16 16-20 20-24 24-28 28-32 32-36
Motomco 919 X -0.86 -0.35 -0.51 1.74*
s 1.02 1.48 1.31 (University of Illinois 1981 Corn)
Dickey-John X 698 715 685 665 656 656
GAC-II X -0.33 -0.10 -0.43 -3.71t
s 0.71 0.74 0.52 o - 30 25 23 21 25 22
N 43 116 152 148 89 50
Burrows 700 X -0.25 0.21 -0.46 -2.04*
s 0.86 0.79 1.01 (Illinois Department of Agriculture 1981 Corn)
SteinUte X — 740 710 686 668 660
SS-250 -0.09 -0.10 0.01 0.07 o - — 9 15 16 26 12
s 0.62 0.56 0.62 N — 3 40 71 48 16
DM X -0.03 -0.28 0.25 1.42
s 0.65 0.68 0.79 (University of Illinois 1980 Corn)
RCT X -0.92 -0.33 -0.59 -3.78t X 765 754 735 696 668 665 646
s 0.71 0.61 0.68 o 7 42 30 17 27 35 40
N 4 22 60 40 18 14 8
Oven X 26.36 25.08 1.28
moisture, % s 1.87 1.95 (Illinois Department of Agriculture 1980 Corn)
Test Weight,
kg/m3
X 692.00
13.00
689.00
12.00
3.0 — X 766 754 716 695 675 677
s o - 24 19 16 18 18 38
* Significant at the 90-9h level. N 21 39 61 58 28 4

rejected and we conclude there was a significant difference between meter


readings for hand-shelled and combine-shelled com.
Also corn tested in 1980 was higher in test weight at all
the Q; = 0.10 level on hand-shelled corn than on combine-shelled moisture ranges than in 1981. Typically, high test weight corn,
corn. For the Steinlite models SS-250 and DM, no significant above 720 kg/m^ (56 Ib/bu), has a test weight factor
difference was found between readings for hand-shelled and subtracted from the moisture reading. Since ILDA samples were
combine-shelled corn. Thus, a partial explanation for the high in test weight, possibly due to hand-shelling, the subtracted
difference in moisture meter bias found between the two test weight corrections could help explain why the ILDA meters
laboratories is that the ILDA samples were all hand-shelled. read drier than the UI meters. [However, this reasoning would
The F statistic was calculated as described in Paulsen et al. not apply to the Motomco 919 meters that use a tapered cell
(1983) to test for differences in the regression coefficients for or the Burrows 700 meters which do not use test weight
the models that predict UI and ILDA meter bias. As a result correction.]
of these tests, we rejected the null hypothesis that the regression
coefficients for the ILDA samples and the UI samples from all Regional Differences
locations are identical. A test to determine which coefficients Tests for possible differences in the regression models
on a particular meter were different, was conducted with the predicting meter performance for samples from different regions
formula: were conducted on the 211, 108, and 52 combine-shelled samples
obtained from lUinois, Indiana, and Ohio, respectively. The
(^:I L ^ I L D A ) / V S IL "^ ^ I L D A Indiana and Ohio samples were collected locally from one
elevator in each state. The IlUnois samples were from many
where: a^^, aj^oA are the intercept coefficients, a, given in locations within the state, but were predominantly form East
Tables 6 and 7. Sj^, ^H^DA ^^^ ^^^ standard errors for the a Central Illinois.
coefficents given in Tables 6 and 7. Slope coefficients and their Tests for comparing the regression coefficients on a state-
associated standard errors were tested in the same manner. The to-state basis for each meter, (using the coefficients and standard
calculated values were greater than 1.96 for both a and b errors given in Table 6) indicate that for th Motomco 919,
coefficients for all six meters. Thus, we concluded that a Burrows 700 and Dickey-John GAC-II, both a and b coefficients
significant difference between the regression models predicting were significantly different between each of the three states. For
meter performance existed between laboratories. the Steinlite SS-250 and RCT the a and b coefficients were
It can not be completely explained why a difference exists, significantly different between IHinois and Indiana, between
but one possible cause already shown is that for certain meters Illinois and Ohio, but not between Indiana and Ohio. For the
hand-shelled corn tends to read significantly wetter than Steinlite DM, the a coefficient did not vary significantly among
combine-shelled corn. Test weight differences and the states; the b coefficient was significantly different between
geographical location, which affect variety and growing IlUnois and Ohio, but not between Illinois and Indiana or
condition, from which samples were taken may also have between Ohio and Indiana, thus, some of the models predicting
contributed to the differences found. meter performance varied by regions. This difference could be
attributed to any number of factors such as corn varietal
Test Weight differences, growing season differences or other factors inherent
Test weight is one of the factors used to adjust moisture meter in samples from that region.
readings. Table 8 indicates that the 20 hand-shelled samples
averaged 692 kg/m^ (53.8 Ib/bu) while the combine shelled Meter Tolerances
corn was 689 kg/m^ (53.6 Ib/bu). Similarly, Table 9 shows that Moisture meter tolerances are illustrated in Figs. 3 through
the ILDA samples that were all hand-shelled had higher test 8 by the *'x" symbols. The " x " symbols denote the 95-percent
weights in both 1980 and 1981 than the UI samples of which confidence limits about the GLS predicted Hne. Confidence
only 33 and 20% were hand-shelled in 1980 and 1981, limits fluctuate at different moisture levels because the variances
respectively. from which the confidence limits were calculated, also fluctuate

1922 TRANSACTIONS of the ASAE—1984


between the 4-percentage point moisture intervals. Typically less than ±0.4 percentage points based on UI results shown
confidence Hmits increase as moisture increases. For example, in Table 10.
for the UI Motomco 919 meter, the confidence limits were ± In comparing estimated cahbration changes (Table 2) to the
0.8 percentage points at 12 to 16% moisture; they were ±3.2 actual changes above 22% moisture (Table 10), all of the meters
percentage points at 28 to 32<7o moisture (Fig. 3). Other models read higher in 1981 than in 1980 except the UI DM and the
had similar confidence Hmits. ILDA RC meters. Since the DM change was a minor one, based
only on test weight changes, it is probable that year-to-year crop
Effect of the Calibration Changes variations were greater than the estimated calibration change.
In the 28 to 32% moisture range the Motomco 919 moisture Within the variability expected due to different crop years,
meter had a positive bias (+ 1.3) for UI samples and a negative varieties, test weights, shelHng methods, and sources of samples,
bias (-0.62) for ILDA samples in 1981. In 1980, at 28 to 32% the calibration changes were within the hmits expected.
moisture, the Motomco had a negative bias ( - 1.65) for the UI
samples and a negative bias (-2.86) for the ILDA samples. CONCLUSIONS
Thus, the Motomco calibration at 30% moisture now reads 2.95 Significant differences in moisture meter performance exists
or 2.24 percentage points higher on 1981 corn than on 1980 corn between hand-shelled and combine-shelled corn samples,
for the UI or ILDA samples, respectively. The actual chart causing certain meters to read lower on hand-shelled samples
change was +2.5 percentage points (Table 2). Thus, the than on combine-shelled samples; thus the relative percentages
accuracy of the Motomco moisture meter was improved by the of hand-shelled to combine-shelled samples should be specified
cahbration changes. in moisture meter cahbrations.
By comparing the 1981 plots shown in Figs. 3 through 8 to Ninety-five percent confidence limits for moisture meter
the similar plots for 1980 data shown in Figs. 3 through 8 in tolerances typically range from ±0.8 percentage points in the
Paulsen et al. (1983), the relative change in meter biases above 12 to 16<^o moisture range to ±3.2 percentage points in the 28
22% can be readily seen. The Motomco meter now reads wetter to 32% moisture range.
for moistures above 22% than it did in 1980, and the change The regression models predicting moisture meter performance
reduced the previous meter bias (Table 10). The Dickey-john vary significantly among samples collected from Illinois,
GAC-II now also reads wetter for moistures above 22% than Indiana, and Ohio, but may not be significant from a practical
it did in 1980, and has reduced bias based on the ILDA results. standpoint because of the large overall sample variability.
For the UI tests the GAC-II bias switched from negative to Differences in moisture meter bias among different samples
positive. The cahbration changes for many of the other meters tested with like meters by the UI and ILDA laboratories, may
had Httle effect on reducing meter bias because in many cases be partially due to differences in hand and combine-shelled
the bias measured during Fall 1980 was already quite low. It samples.
should be noted that the Steinlite meters had been recahbrated Motomco moisture meter accuracy at moistures above 22%
prior to the Fall 1980 tests and during Fall 1980 their bias was is improved by the 1981 calibration changes.
Variability in measuring moisture meter bias increases as oven
moisture increases.
TABLE 10. MOISTURE METER BIAS BEFORE AND AFTER
THE FALL 1981 CALIBRATION CHANGES FOR ALL SAMPLES References
TESTED IN 1980 AND IN 1981
1. ASAE Standard: ASAE S352. Moisture measurement - grain and
Below 22% moisture Above 22% moisture seeds. 1983. Agricultural Engineers Yearbook.
2. Hurburgh, Jr., C. R., C. J. Bern, and S. N. Grama. 1981.
Meter model FaU 1980 FaU 1981 Fall 1980 FaU 1981 Improvements in the accuracy of corn moisture measurement in Iowa.
(University of Illinois) ASAE Paper No. 81-3515, ASAE, ST. Joseph, MI 49085.
Motomco 919 0.06 -0.18 -0.97 0.53 3. Paulsen, M. R., L. D. Hill, and B. L. Dixon. 1983. Moisture meter-
Dickey-john to-oven-comparisons for Illinois corn. TRANSACTIONS of the ASAE
GAC-II 0.33 0.00 -0.83 1.10 26(2):576-583.
Burrows 700 — -0.09 — 0.68 4. Paulsen, M. R., L. D. Hill, and B. L. Dixon. 1982. Moisture meter-
Steinlite to-oven comparisons for Fall-1981 corn. ASAE Paper No. 82-3548,
SS-250 0.38 0.04 -0.36 0.41 ASAE, St. Joseph, MI 49085.
DM 0.15 -0.29 -0.12 -0.48
Automatic 0.21 0.12 5. SAS User's Guide, 1982. SAS Institute Inc., Raleigh, NC.
— — 6. Schwarberg, R. L., S. E. McGuire, and J. Martin. 1982.
RCT 0.03 -0.14 -0.17 0.54
Comparison of moisture of the 1981 Illinois high moisture corn crop
(lUinois Department of Agriculture) (22-37<7o) vs. the 1980 results by commercial moisture meters using the
Motomco 919 -0.17 0.05 -1.84 -0.70 official 72 hour air oven method as a standard. Illinois Department
Dickey-john of Agriculture, Bureau of Laboratories, Springfield, IL.
GAC-II 0.29 0.54 -0.99 -0.12 7. Snedecor, G. W. and W. G. Cochran. 197L Statistical methods.
Burrows 700 0.12 0.88 -1.01 -0.47 6th ed., The Iowa State University Press, Ames, lA.
Steinlite 8. USDA Equipment Manual. Insert in GR Instruction 916-6. 1971.
SS-250 0.33 0.51 -0.49 -0.41 USDA, Consumer and Marketing Service, Grain Division, Hyattsville,
DM 0.02 0.55 -0.80 -0.55 MD 20782.
Automatic 0.22 0.57 -0.33 -0.55 9. Wollin, H. F., L. E. Barbrow, and A. P. Heffernan. NBS Special
RCT 0.06 0.76 -0.85 -0.69
RC -0.09 0.55 -1.17 -1.04 Publication 629. Report of the 66th National Conference on Weights
DL 0.44 1.19 -0.58 -0.06 and Measures 1981. National Bureau of Standards, January 1982.
G-400 0.39 1.33 0.28 0.56 Illinois-Iowa Moisture Meter Task Force, Lowell Hill reporting, p.
224-234.

1984—TRANSACTIONS of the ASAE 1923

You might also like