1989)
215 USE OF NEAR INFRARED ANALYSIS FOR T H E N O N D E S T R U C T I V E M E A S U R E M E N T OF DRY M A T T E R IN POTATOES Gerald G. DulP, Gerald S. Birth 1, and Richard G. Leffler1 Abstract
Near infrared spectrophotometry in the direct transmittance mode and in the wavelength range of 800 to 1000 nm was used to measure the % dry matter (DM) in sliced and intact potato tubers (cv Russet Burbank). With thin tissue slices, the correlation (r) between spectral and D M data was -0.975 with a standard error of calibration (SEC) of 0.91. With thick slices and intact potatoes the r, SEC, and standard error of prediction (SEP) values were -0.952, 1.28, 1.69 and -0.918, 1.04 and 1.52, respectively.
Compendio Se utiliz6 la espectrofotometria cercana al infrarojo en transmisi6n directa y e n intervalos de longitud de onda de 800 a 1 000 nm, para medir el yo de materia seca (MS) en tub6rculos de papa cortados en rebanadas, y enteros (cv Russet Burbank). Con rebanadas delgadas de tejido la correlaci6n (r) entre los datos del espectro y la MS fue de -0,975 con un error estfindar de calibraci6n (EEC) de 0,91. Con rebanadas gruesas y tub6rculos enteros los valores de la r, EEC, y el error est~indar de predicci6n (EEP) fueron -0,952, 1,28, 1,69 y -0,918, 1,04, y 1,52, respectivamente.
Introduction In the production and processing of potatoes (Solanum tuberosum L.), quality and yield of processed products are directly correlated with specific gravity (SG) and % dry matter (DM), making these measurements important indices of quality (5, 7). It should be noted that other important quality parameters for potatoes and potato products include defects, sugar levels, and color. Specific gravity is routinely measured nondestructively by weighing; however, the method is at least an order of magnitude slower than the spectrophotometric determination of other quality parameters (2). In a typical procedure, a 5 kg sample of tubers is weighed in air and in water and the SG is calculated. The importance of this analytical method is further 1United States Department of Agriculture, Agricultural Research Service, R.B. Russell Agricultural Research Center, Athens, Georgia 30613. Accepted for publication December 7, 1988. ADDITIONAL KEY WORDS: Potato, dry matter, specific gravity, NIR, spectrophotometry.
216
AMERICAN POTATOJOURNAL
(Vol. 66
indicated by the fact that an automated, computer controlled device has been developed for measuring 8G (11). Dry matter is an important, although not necessarily the most limiting, quality parameter for potato products. 8immonds (10) has summarized a number of published accounts of the relationship between SG, D M and starch content. In terms of estimating D M and starch content from SG determinations, he notes that "it is evident that the empirical linear regression approach to estimation remains the only feasible one." 8amotus, et al. (8) have thoroughly reviewed factors that contribute to discrepancies in the determination of D M and starch from potato tuber density. Those factors include air absorbed on tuber surfaces, errors in D M determination, and errors in the determination of intercellular spaces. Further, there is an inherent variation in chemical composition among individual potatoes in a single population harvested at the same time from the same field. A group of potatoes characterized by a single 8G value may well include a wide range of individual tuber 8G values. Further, there is a variation in SG with location in individual tubers (12). Considering these variables, the use of SG to estimate the D M in potatoes is at best a rough average. Even with its drawback, specific gravity measurement is presently the best practical and nondestructive method for estimating DM. A much more rapid and accurate nondestructive method for determining D M in individual tubers would appear to have considerable utility for potato research and in the production and processing of potatoes. Near infrared (NIR) spectrophotometry has been used to determine the dry matter in intact onions (2). In this work, the nominal wavelength selected for the regression equation was associated with a broad carbohydrate absorption band at 918 nm. The success of this method is attributed to the fact that over 80% of the dry matter in onions is carbohydrate. Since a similar situation exists for white potatoes (Solanum tuberosum) (4), this study was undertaken to evaluate the possibility of using N I R for the rapid and nondestructive determination of dry matter in individual potato tubers. Materials and Methods
Instrumentation and Spectral Data Acquisition Two high intensity, single beam N I R spectrophotometers, a biological spectrophotometer (the Biospect) (3), and the LT 7000 (LT Industries, Rockville, MD) were used to collect spectral data. Both instruments are computer controlled. Spectral measurements on slices and intact tubers were made by direct transmittance. This is an arrangement where the radiation enters the sample at one area, is transmitted through the sample, and is then monitored by a detector positioned on the opposite side of the sample. The reference data for direct transmittance were taken using a neutral density filter in place of the potato. The ratio of the data for the potato to the reference data samples yielded a transmittance (T) spectrum
DULL, eta/." INFRAREDANALYSIS
1989)
217
for the potatoes that was converted to an optical density spectrum by taking the logarithm of 1/T. Samples were scanned from 800 to 1000 nm.
Data Processing Data processing and regression computations were carried out with a single computer program as described by Norris and Massie (6). Data processing consisted of the computation of second derivatives of the optical density spectral data. A linear regression analysis program was used to calculate ratios of second derivatives at all possible wavelengths. Adjustable parameters in the calculation of the second derivative ratios included the number of adjacent points to average, the gaps between point groups for the numerator and the denominator, and the numerator and denominator wavelengths. The objective was to find second derivatives centered at two different wavelengths such that their ratio gave the highest correlation coefficient when regressed with the % DM. The use of a second derivative ratio appears to eliminate sample size and light scattering as factors affecting the quality of the prediction. This procedure is detailed by Birth, etal. (2). The regression equation arising from this analysis is: % D M = K0 + K~ X where K0 is the regression equation intercept K1 is the regression equation slope X is the ratio of two second derivatives A regression equation developed with a calibration data set is referred to as a calibration equation. The predictive quality of the relationship between spectral and D M data was evaluated using the correlation coefficient (r), the standard error of calibration (SEC) and the standard error of prediction (SEP). The SEP was obtained when a calibration equation from one data set was used to predict the D M in a different data set.
Samples Russet Burbank potatoes were supplied by the Lamb Weston Company as lots separated into consecutive 0.005 SG unit categories. Lamb Weston separated the categories by a brining procedure. The amount of potatoes in a lot varied from 10 to 50 kg, depending upon the availability. The lots ranged from 1.060 to 1.100 SG. For an experiment, the appropriate number (the number varied with the experiment) of potatoes was selected from each SG lot and the SG was determined for each potato.
Specific Gravity Measurements Individual slices and intact potatoes were weighed in air and water to the nearest 0.01 g. The potatoes and water were allowed to reach the
218
AMERICAN POTATO JOURNAL
(Vol. 66
ambient temperature before weights were taken. T h e SG was calculated by the formula: weight in air SG = weight in air--weight in water
Dry Matter Determination Samples weighing less than 10 g were dried in a vacuum oven at 60 C for 16 hours. Larger samples were dried to a constant weight (approximately 48 hours) in a freeze drier. T h e two methods gave the same results within experimental error.
Specific Gravity Measurements Within a Potato Tuber Using a meat slicer, a single potato was cut into 5 mm slices, perpendicularly to the long axis and from the stem end (slice 1 ) to the bud end (slice 17). T h e specific gravity was determined for each slice.
Dry Matter Measurements Within a Potato Tuber Potatoes were selected from the 1.060, 1.080 and 1.100 SG lots and cored (2.54 cm diameter) perpendicular to the long axis at the stem end, center and bud end. T h e cores were cut into circular slices (5 m m thick) and the D M in each slice was determined by vacuum oven drying.
Thin Slice Preparation Six potatoes were selected to cover as broad a D M range as possible. Multiple 2.54 cm cores were taken from each potato. Each core was sliced into 5 m m thick consecutive slices that were scanned 100 times in the L T 7000 spectrophotometer and averaged. After spectral measurements were taken, each slice was vacuum oven dried to obtain the % D M .
Thick Slice Preparation Two lots of 10 potatoes each were selected to cover as wide a range of D M as possible. Longitudinal slices were removed from opposite sides of each tuber to produce a remaining thick slice with parallel faces. T h e 20 thick slices ranged from 4 to 6 cm in thickness. Each sample was scanned in the Biospect at four different spots on the top and bottom, giving 8 separate spectral scans per thick slice. T h e Biospect was used in order to obtain an adequate signal through the thick samples. Following spectral analysis, the samples were diced and dried in a freeze drier to obtain a single D M value for the whole sample. One of the 10 tuber sample lots was used to develop a calibration data set and data from the other were used for prediction.
1989)
DULL, et aL" INFRAREDANALYSIS
219
lntact Potatoes Two lots of 30 potatoes each were selected to cover the SG range of 1.060 to 1.100. In the NIR spectrophotometric method it is important that the samples contain as broad a concentration range of the constituent being determined as possible. This contributes to a lower SEC. Using the Biospect, each sample was scanned with the radiation traversing the potato through its narrowest dimension at the center. The potato was rotated 180 ~ about its long axis and another spectrum was taken. Our experience indicates this procedure is very important for achieving a realistic average over the potato tissue. This procedure produced 60 separate spectra for each lot. One lot of potatoes was used to develop a calibration data set and the second lot was used for prediction.
Results and D i s c u s s i o n Synopsis o/the Approach To establish the validity of the use of NIR spectrophotometry for the quantitative determination of DM in white potatoes, five separate experiments were done: 1. A determination was made of how SG changes along the length of a typical potato; 2. The establishment of how DM varies throughout a potato so that an optimal light path through the potato could be chosen; 3. The determination of D M in thin potato slices as an ideal geometry for the method; 4. The determination of DM in thick slices to evaluate the method without interference from the potato skin; and 5. The determination of DM in intact potatoes.
Specific Gravity Variations Within a Potato Tuber In an experiment designed to provide information for developing an acceptable spectrophotometric sampling procedure, the SG of consecutive slices perpendicular to the long axis in a tuber was determined. Tlae results presented in Table 1 clearly indicate a large variation ( 1.048 to 1.076) in the SG of tissue within a tuber and are consistent with the findings of Whittenberger and Nutting (12).
The Dry Matter Distribution Within a Potato Tuber Three tubers representing low, medium and high SG (and therefore DM content) were sampled to provide cores perpendicular to the long axis. Each core was cut into slices and the DM for each slice determined by vacuum oven drying. The results are presented in Table 2. The ranges of DM within single tubers were 14.1 - 21.8, 17.3 - 26.0, and 16.8 - 32.5%. These results are in agreement with those of Whittenberger and Nutting (12). These ranges illustrate a potential difficulty of sampling a tuber with transmitted NIR radiation to estimate the DM for the entire tuber. It is clear
220
AMERICAN POTATO JOURNAL
(Vol, 66
TABLE 1. - - Specific gravity of consecutive slices perpendt~lar to the long axis
in a single white potato Sample No.
Specific Gravity
Sample No.
Specific Gravity
1a 2 3 4 5 6 7 8 9
1.048 1.066 1.068 1.076 1.076 1.072 1.073 1.074 1.071
10 11 12 13 14 15 16 17
1.070 1.076 1.076
1.075 1.075 1.073 1.073 1.066
a Bud end.
TABLE 2. -- Dry matter distribution in three white potatoes. % Dry Matter Pot. No. 1
2
3
Core No. a
Slice Number b 1
2
3
4
5
6
19.1 19.1 18.8
18.7 17.1 17.3
17.1 14.1 15.2
15.5
2 3
17.1 17.4 21.8
17.9 18.4
1 2 3
23.3 25.1 25.2
27.8 23.7 23.0
20.6 19.9 19.8
22.7 24.1
26.0
21.1 22.2 25.3
1 2 3
28.8 24.5 23.6
23.0 19.7 25.3
18.7 16.8
21.9 20.7 22.9
24.6 26.1 24.1
28.5 30.4 24.0
1
25.5
14.9 17.3 17.3
17.7
7
32.5 22.5 19.5
a Core 1 is from stem end; core 2 from center; core 3 from bud end. b First and last slices in a core represent periderm tissue.
t h a t t h e c e n t e r ( p i t h ) of t h e t u b e r has less d r y m a t t e r t h a n t h e o u t e r tissue (cortex). T h e r e is also a v a r i a t i o n in t h e average D M in t h e t h r e e cores f r o m each t u b e r . T h e c o r r e l a t i o n b e t w e e n c o r e D M a n d w h o l e p o t a t o D M was h i g h e s t for t h e c e n t e r core. Since t h e p u r p o s e of this w o r k is to m a k e a r a p i d a n d n o n d e s t r u c t i v e m e a s u r e m e n t of t h e average D M of a single i n t a c t p o t a t o , it was c o n c l u d e d t h a t a t r a n s v e r s e l i g h t p a t h t h r o u g h t h e w h o l e p o t a t o at t h e c e n t e r p o i n t of t h e l o n g axis p r o v i d e d t h e m o s t r e p r e s e n t a t i v e single s a m p l i n g of t h e w h o l e p o t a t o . T h i s m e a s u r e m e n t s t r a t e g y was u s e d s u b s e q u e n t l y on i n t a c t p o t a t o e s .
DULL, et al: INFRARED ANALYSIS
1989)
221
Thin Slices As a first step in establishing the validity of using spectrophotometric data for the estimation of D M , thin slices of potatoes were scanned in the LT 7000 instrument. This eliminated the skin as a source of variation and optimized the signal/noise ratio. T h e relevant results for the linear regression analysis of the spectral and D M data are given in Table 3. T h e actual D M values ranged from 14.0 to 33.2%. T h e correlation coefficient (r) for the relationship between spectral and D M data was -0.9749 with an SIfC of 0.91. With a sample size of 52, it is not likely that the high correlation is due to chance (1). Another way of evaluating the quality of the relationship between spectral and chemical data is to use the calibration equation developed with a data set to predict the same data set. A plot of the predicted D M values vs the actual D M values for the thin slice calibration data set is shown in Figure 1. A prediction data set was not obtained with the thin slices since the purpose of the experiment was simply to establish the quality of the correlation between spectral and D M data.
Thick Slices Ten thick slice samples were scanned in the Biospect at four different spots on the top and bottom to obtain a total of 80 individual spectra. T h e regression equation parameters for this analysis are given in Table 3. Figure 2 shows a plot of the predicted D M values vs the actual D M values for the thick slice prediction data set. T h e r e is good agreement between the results for thin and thick slice measurements even though the n u m b e r of potatoes in the sample was lower for the thick slice experiment. T h e r is slightly lower
TABLE
3.
--
Regression equation parameters f i r N I R determination of % D M in sliced and intactpotatoes.
Wavelength Sample
Regression
(nm) Num 907
Den 876
Nz 52
KoY 47.9
Kl -13.6
rx SEC w 0.9749 0.91
Slices (thin) Slices 907 870 80 43.3 - 9.9 -0.9520 (thick) Intact 910 881 60 35.9 - 5.4 -0.9178 tubers ZN = Number of samples used in the regression analysis YK0 and KI = regressioncoefficients x r = linear correlation coefficient w SEC = standard error of calibration v SEP = standard error of prediction
8EP v
1.28
1.69
1.04
1.52
222
AMERICANPOTATOJOURNAL
(Vol. 66
32. o
28
o o
24-
o
s
D
o o
lid 0
0 O
20.
O 0
O
16,
12
'
2
I
16
~
I
20
i
I
24
'
I
28
'
I
32
ACTUAL DRY MATTER (%) FIG. 1. A plot of predicted DM vs actual DM in thin potatoslicesin the calibrationdata set. and the SEC is higher for the thick slices compared with thin slices. These differences are attributed to the reduced signal at the detector for thick slices. Intact Potatoes
In this experiment, 30 potatoes were scanned twice, producing 60 spectra. There was a single DM value for each potato. The spectral and D M data for the calibration samples were subjected to linear regression analysis as described above and the regression equation parameters are listed in Table 3. The wavelengths selected by the regression analysis~program, 910 and 881 nm, agree well with those selected for the thin and thick slices, indicating that the introduction of the skin into the light path does not alter the basic relationship between the spectral and D M data. The SEC was close to that for the thin slices, indicating that the predictive capability of the method holds for intact potatoes as well as slices. The true test of the method is to use the calibration equation to predict the DM in a different potato sample set, and the second lot of 30 intact potatoes was used for this purpose. This resulted in an SEP of 1.52. In the
1989)
DULL, et al.: INFRAREDANALYSIS
223
28o
24-
~
20-
o
o o
~0
I
I
f~ 16.
12
i
12
1
;
16
20
ACTUAL DRY MATTER
l
=
24
l
28
(%)
FIG. 2. A plot of predicted DM vs actualDM in thick potatoslicesin the predictiondataset.
onion work (2}, markedly improved results were obtained when two separate predictions were made for one sample and those two values averaged to give a single predicted DM value. This procedure was followed in the present work and the SEP was reduced to 1.28. Figure 3 is a plot of the predicted D M vs the actual DM for intact potatoes in the prediction set. In a six year study, Schippers (9) investigated the relationship between SG and D M in potatoes. In those studies, which covered 18 cvs, 5 crop years and over 270 samples, the correlation coefficients and SEC ranged from 0.915 to 0.956 and 2.00 to 2.88, respectively. It would be expected that the SEP range would be greater than the SEC range. These results indicate that the precision of the NIR method is equal to or better than that for the SG method for estimating dry matter in intact potatoes. The value of our method is that, given a calibration for a potato population, the basic spectrophotometric estimation can be accomplished in less than 1 second. Because of the inherent variation of the DM within a single tuber, additional work will be necessary to establish the optimal illumination pattern in a tuber and the minimum number of spectral measurements needed to achieve a given level of predictability.
224
AMERICANPOTATOJOURNAL
(Vol. 66
28D
,,=, 0
0
0
D
0
o
a
0
0
20
O0
0
0 0
0
0
16 16
20
24
28
ACTUAL DRY MATTER (%) FIG. 3. A plot of predicted DM vs actual DM in intact potatoesin the prediction data set.
Summary The use of N I R spectrophotometry for the prediction of D M in potatoes gave r values of-0.9749, -0.9520 and 0.9178 for thin slices, thick slices, and intact tubers, respectively. These are high correlations when one considers the complexity of the physical and chemical nature of the samples. With the large sample numbers, these correlations are not likely due to chance. The similarity of the regression parameters achieved for the potato results to those for the N I R determination of dry matter in onions supports the validity of the method. An improvement in the precision of the method seems likely with the use of instrumentation having a more intense radiation source and more sensitive detectors. A multiple regress'ion data analysis approach might also improve the precision of the method.
Acknowledgments The technical assistance of Carol Walter, Judy Brewer Arthur and Alice Wilcher is gratefully acknowledged. The cooperation of the Lamb Weston Company in supplying the potato samples is deeply appreciated.
1989)
DULL, et al." INFRARED ANALYSIS
225
T h e m e n t i o n of a specific p r o d u c t o r a t r a d e n a m e does n o t i m p l y e n d o r s e m e n t b y t h e U n i t e d States D e p a r t m e n t of A g r i c u l t u r e to t h e exclusion of s i m i l a r p r o d u c t s of e q u a l q u a l i t y o r p e r f o r m a n c e .
Literature Cited
1. Birth, G.S. 1985. Evaluation of correlation coefficients obtained with a stepwise regression analysis. Applied 8pectr 39 (4):729-732. 2. Birth, G.S., G.G. Dull, W.T. Renfroe and S.J. Kays. 1985. Nondestructive spectrophotometric determination of dry matter in onions. J Am Soc Hortic Sci 110 (2):297-303. 3. Birth, G.S. and G.L. Zachariah. 1973. Spectrophotometer for biological applications. Trans Am Soc Agric Eng 16 (2):371-373. 4. Brautlecht, C.A. and A.S. GetchelL 1951. The chemical composition of white potatoes. Am Potato J 28:531-550. 5. Lulai, E.C. 1986. Potato specific gravity. Chipper/Snacker 43 (8):28-29. 6. Norris, K.H. and D.R. Massie. 1981. Multipurpose, linear regression program for analysis of spectrophotometric data. Proc 1981 Conf of HP 1000 Int Users Group, Washington DC, Dec. 2. 7. Orr, P.H. and C.K. Graham. 1983. A generalized model for determining least-cost sources of processing potatoes. Tram Am 8oc Agric Eng 26 (6): 1875-1878. 8. Samotus, B., M. Leja, A. Scigalski, J. Dulinski and R. Siwanowicz. 1986. Analytical remarks to explain some discrepancies in the determination of dry matter and starch from potato tubers' density. Starch 38:14-19. 9. Schippers, P.A. 1976. The relationship between specific gravity and percentage dry matter in potato tubers. Am Potato J 53" 111-112. 10. 8immonds, N.W. 1977. Relations between specific gravity, dry matter content and starch content of potatoes. Potato Res 20:137-140. 11. Tai, G.C.C., G.C. Misener, E.8. Allaby and L.P. McMillan. 1985. Gray-O-Tater: A computer apparatus for measuring specific gravity. Am Potato J 62:403-408. 12. Whittenberger, R.T. and G.C. Nutting. 1950. Observations on sloughing of potatoes. Food Res 15:331-339.