Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/258101934
CITATIONS READS
26 1,391
4 authors, including:
Joo A Lopes
University of Lisbon
110 PUBLICATIONS 1,239 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Cludia Santos on 19 March 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
focal point review
CLAUDIA A. TEIXEIRA DOS SANTOS, MIGUEL LOPO,
RICARDO N.M.J. PASCOA, JOAO A. LOPES
UNIVERSIDADE DO PORTO, REQUIMTE,
DEPARTAMENTO DE CIENCIAS QUIMICAS, FACULDADE DE FARMACIA,
RUA DE JORGE VITERBO FERREIRA, 228, 4050-313 PORTO, PORTUGAL
T
internal and external parameters of fruits and he agro-food industry has a
moisture, protein, and fat content of a
vegetables; conservation state and fat content significant role in the world
wide range of agricultural and food
of meat and fish; distinguishing among and economy. In the last decades, it
products.7,8 Early applications of this
quality evaluation of beverages and dairy has followed this revolution to meet technique included determinations of
products; protein content of cereals; evaluation increasing consumer needs and expecta-
of grape ripeness in vineyards; and soil
dry matter content in onions,9 soluble
tions concerning quality, convenience, solids content (SSC) in apples,10 and
analysis. Chemometrics is an essential part of diversity, and health. Simultaneously,
NIR spectroscopy manipulation because wave- water content in mushrooms.11 This
there is a need to meet health and safety technique was also used to perform
length-dependent scattering effects, instrumen-
tal noise, ambient effects, and other sources of requirements; increasingly demanding microstructural measurements of stiff-
variability may complicate the spectra. As a environmental legislation; and security, ness,12 internal damage,13,14 and even
consequence, it is difficult to assign specific ethics, and sustainable production re- sensory attributes.15 Numerous studies
absorption bands to specific functional groups. quirements. The quality of agricultural have been published involving NIR
To achieve useful and meaningful results, and food products is classified according spectroscopy and its usefulness in the
multivariate statistical techniques (essentially to internal and external parameters.1 agro-food industry.2,1618
However, several analytical techniques Recently, considerable attention has
Received 19 July 2013; accepted 27 August have been applied to assess those been given to the miniaturization and
2013. parameters; the invasive, destructive, portability of spectroscopic devices.19
* Author to whom correspondence should be
sent. E-mail: joaolopes@ff.up.pt. and time-consuming nature of the refer- The use of portable instruments for in-
DOI: 10.1366/13-07228 ence methods has spurred efforts toward field measurements offers interesting
Commercial Resolution
name Manufacturer Models available Spectral range (nm) (nm) Light source Signal to noise
AgriSpec ASD Not applicable 3502500 3 at 700; 10 at Halogen 14 500 S/N at 1400 nm
1400/2100
FieldSpec4 ASD a. FieldSpec4 Standard- a, b, c, 3502500 a, d, 3 at 700; 10 Halogen Not specified
Res d, 3251075 at 1400/2100
b. FieldSpec4 Hi-Res b, 3 at 700; 8 at
c. FieldSpec4 Wide-Res 1400/2100
d. FieldSpec Handheld 2 c, 3 at 700; 30 at
1400/2100
LabSpec4 ASD e. LabSpec4 Standard- 3502500 e, 3 at 700; 10 at Halogen Vis-NIR 9000:1 at 700 nm
Res 1400/2100 SWIR 1 9000:1 at 1400 nm
f. LabSpec4 Hi-Res f, 3 at 700; 6 at SWIR 2 4000:1 at 2100 nm
1400/2100
AvaSpec Avantes Not applicable 2001100 0.0520 Not specified 200:1
Luminar 5030 Brimrose Not applicable 6001100, 8501700, 210 Not specified Not specified
9001800, 11002300
microPHAZIR Thermo Not applicable 16002400 Not specified Tungsten light bulb Not specified
Scientific
Phazir Polychromix g. DTS-Phazir-1016 g, 10001600 Not specified Tungsten light bulb g, 8 nm
h. DTS-Phazir-1624 h, 16002400 h, 12 nm
USB Ocean Optics USB 4000 VIS-NIR Grating dependent ~1.5 Tungsten light bulb 300:1 (at full signal)
a
AC/DC, alternating current/direct current; SWIR, short wave infrared; USB universal serial bus.
possibilities by allowing direct measure- broad and superimposed.22 The funda- measurement, and high performance
ment of important parameters. These mentals of vibrational spectroscopy reliability. Another important character-
devices have several advantages, includ- provide the theory necessary to interpret istic to be taken into account is instru-
ing reduced cost, increased safety for the this characteristic spectrum.23 The typi- ment design. An ergonomic design
environment and operators, and the cally observed bands in NIR spectra should make these portable devices
improved accuracy of the method by correspond to bonds containing the handheld, thus enhancing their ease of
preserving the sample and avoiding its hydrogen atom, such as CH, NH, use. While some devices are sold as
transport. In recent years, several efforts OH, and SH, that are frequently closed instruments (and cannot be
have been made to apply this technique present in most organic and some adapted), others offer the possibility of
to solid-sample analysis. Salvador Gar- inorganic compounds.17 A NIR spec- interchanging measurement accessories
rigues and Miguel de la Guardia recently trum is essentially composed of over- so that the same instrument can be used
summarized studies using portable in- tones and combination bands containing for different sample requirements.25
struments and their wide use in a vast useful chemical and even physical Several portable NIR spectrometers are
field of applications.2 information. The literature on NIR currently available in the marketplace,
This study provides an overview of spectroscopy and its applications in varying in cost and purpose. Table I lists
the more recent applications of portable different fields is extensive and can be some of the most important specifica-
NIR instruments in the agro-food sector. found elsewhere.17 tions of the portable NIR instruments
These applications have been divided Portable Near-Infrared Spectrom- most commonly found in the literature
into fruits and vegetables, meat and fish, eters. Industrial applications created the with reported applications in the agro-
beverages and dairy, cereals and feed- need for a cost-effective and nonde- food industry.
stock, soils and manures, and viticulture structive quality-control system. This Near-Infrared Spectral Processing.
and olive growing. Descriptions of the requirement increased the interest in Chemometrics, usually described as the
applications are provided, highlighting NIR spectroscopy, leading to the trans- application of mathematics, multivari-
the potential of this method in terms of formation of large, stationary analytical ate statistics, and computer science,18
its versatility. instruments into lightweight tools that plays a fundamental role in the extrac-
enabled new applications to be imple- tion of all-important relevant informa-
NEAR-INFRARED mented in situ at an industrial level.24 tion hidden in NIR spectra. Because a
SPECTROSCOPY Although these portable NIR spectrom- NIR spectrum consists of weak and
Near-infrared radiation ranges from eters offer several advantages, such as broad absorption bands, a straightfor-
780 to 2500 nm of the electromagnetic nondestructive and in situ analyses, their ward interpretation is often not possi-
spectrum.20,21 The measurement of the development must consider some critical ble. Furthermore, there are several other
interaction between this radiation and a factors, such as cost, size, weight, power factors that increase the complexity of
sample results in a spectrum character- consumption, robustness, safety, user- the spectra: water absorption bands,
ized by weak absorption bands that are friendliness, durability, accuracy of scattering effects, instrumental noise,
AgriSpec Customizable 100 Ethernet, RJ-45, AC/DC, battery 040 5.4 127 3 368 3 292
Wifi (2 h)
FieldSpec4 Customizable a, b, c, 100 a, b, c, Ethernet, a, b, c, AC/DC, 040 a, b, c, 5.44 a, b, c, 127 3 368 3
d, Not specified Wifi battery (6 h) d, 1.2 292
d, USB d, AC/DC, battery d, 90 3 140 3 215
(1.55 h)
LabSpec4 Customizable 100 Ethernet, Wifi AC/DC, battery 040 5.44 127 3 368 3 292
AvaSpec Customizable 1.8 USB, RS-232 USB, AC/DC Not specified 0.716 175 3 110 3 44
Luminar 5030 Diffuse reflectance 16 000 Ethernet AC/DC Not specified Not specified Not specified
wavelength/s
microPHAZIR Diffuse reflectance ~1 min USB Battery (3 h) 550 1.8 266 3 109 3 251
Phazir Diffuse reflectance 12 s USB Battery (58 h) 545 1.7 254 3 292 3 152
USB Customizable .5 ms USB AC/DC Not specified 0.190 89.1 3 63.3 3 34.3
sample complexity, and matrix/environ- developed to unlock spectral data and prediction (RMSEP) or root mean
mental effects.16 Therefore, multivari- enhance the robustness and predictive square error of cross-validation
ate analysis methods coupled with ability of the calibration model.22 Che- (RMSECV) as indicators of NIR-based
spectral preprocessing techniques are mometric methods most commonly used model accuracy. The equation for root
the fundamentals of chemometrics for to process NIR spectral information and mean square error (RMSE) is
the appropriate treatment of analytical build calibrations are available in the v
literature and are not developed u 2
information contained in NIR spectra.26 uX in
here.18,32,33 In terms of their purposes u
Essentially, spectral information cap- u y i -^ y i
tured with portable instruments is and their algorithms or computational t i1
procedures, the most commonly used RMSE 1
processed using standard chemometric n
methods that have been thoroughly multivariate analysis methods are listed
described in the literature.27 Reducing in Table II. The results presented here where ^yi is the prediction value of the ith
instrumental noise or background infor- use the root mean square error of observation, yi is the measured value of
mation is usually performed using the observation, and n is the number of
smoothing techniques. A detrending the observations. Equation 1 can be used
TABLE II. Classification of the major quali- to compute the root mean square error of
preprocessing technique is sometimes tative and quantitative multivariate-analysis
used to remove the effects of accumu- calibration (RMSEC), RMSECV, and
methods used in NIR spectroscopy applied to
lating data sets from a trend, so that the agro-food industry.17,a RMSEP.27
only the absolute changes are shown.28
Derivatives are a common method used Quantitative APPLICATIONS IN THE AGRO-
to eliminate unimportant baseline sig- Non-linear methods FOOD INDUSTRY
Linear methods
nals from samples.29 Orthogonal signal ANN, non-linear PLS In the following sections a full
correction (OSC) was developed to MLR, PCR, PLS, LS-SVM description is given of the applications
reduce light-scattering effects, among Qualitative of portable NIR spectrometers in the
other general types of interference (not Unsupervised methods agro-food industry including the type of
Supervised methods
interfering with the correlation with Cluster analysis, ANN (e.g., Kohonen sample and its attributes, and, when
reference values). Path length effects, networks), PCA available, the spectrometer used, its
scattering effects, source or detector FDA, KNN, LDA, SIMCA acquisition mode, and its spectral range
variations, and other general instrument a
FDA, Fisher discriminant analysis; KNN, k- (in most studies the visible spectral
sensitivity effects are usually cor- nearest neighbor; LDA, linear discriminant analy- region was also included). This review
rected using normalization preprocess- sis; LS-SVM, least squares support vector ma- considers the application of portable
ing methods, such as standard normal chines; MLR, multiple linear regression; PCA, NIR instruments in the agro-food indus-
principal component analysis; PCR, principal
variate (SNV)28 and multiplicative sig- component regression; PLS, partial least squares; try independently of the application
nal correction (MSC).30,31 Consequent- SIMCA, soft independent modeling of class conditions. It is clear that the great
ly, preprocessing techniques, have been analogy. advantage of these systems is the
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
possibility of using them under real tential of NIR spectroscopy as applied the goal of predicting whether the fruit
production conditions; however, many to fruit and vegetables16 and about was ready to be harvested. Most
of the studies considered here were enhancing its use as a rapid and studies in this area model parameters
actually performed under laboratory nondestructive analytical technique commonly used as maturity indicators,
conditions (involving sampling and useful for determining internal and such as soluble solids content
transportation of the samples to the external characteristics, either quantita- (SSC),34,35,3741,4365 titratable acidi-
laboratory). Therefore, these studies do tive or qualitatively. However, the fruit ty, 3 7 , 4 0 , 4 8 , 5 1 5 3 , 5 5 , 5 8 , 6 0 , 6 2 , 6 4 , 6 6 and
not actually assess the performance of and vegetable sector demands field pH.55,60,62 External quality parameters
the systems under real production con- NIR spectroscopy equipment capable were also investigated, with special
ditions and are not examples of taking of performing in situ analyses to allow attention given to color;40,44,48,49,62,67
direct in situ measurements under the producers to establish the most suitable weight,44,62 and size.54,62 Among the physi-
uncontrolled conditions typical of pro- harvest dates, consequently ensuring cal parameters considered, firmness was
duction sites. the best possible crop quality, while the most common,42,44,4749,52,54,56,62,67
Fruits and Vegetables. Most of the avoiding waste and loss. Table III with some attention also paid to juice
studies in this area were performed in summarizes portable NIR spectroscopy content and juice weight,62 pericarp
the laboratory, applying NIR spectro- applications to fruit. Promising results thickness,58,62 and maximum penetra-
scopic instruments to intact harvested have been achieved when portable NIR tion force.62 In addition, Beghi et al.52
fruits (in some cases, spectra were spectroscopy has been applied in this and Guidetti et al.34 explored some
collected from the same fruit in more sector, indicating the successful use of nutraceutical parameters: total anthocy-
than one position or using spinning NIR spectroscopic measurements in anin, total phenol, total flavonoid, non-
accessories). There are, however, a nondestructively evaluating and moni- anthocyanic flavonoid, carotenoid, chlo-
significant number of applications in- toring fruit-quality parameters. 3436 rophyll, and ascorbic acid content.
volving the direct use of the equipment The potential of this technique has The correlation of the measured
in the field, where measurements were been highlighted for assessing the chemical and physical parameters with
made directly of on-tree fruits under internal quality of fruits nondestruc- the environmental conditions to which
varying weather conditions. There are tively in the field under different fruits were exposed was often evaluat-
numerous studies highlighting the po- weather conditions,3747 usually with ed. 65,68 Environmental factors are
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Swordfish Correct classification MMS1 (Zeiss), 380 PLS-DA, PCA 96.7% correctly 81
according to the 1080, reflectance classified
preservation status
Pork meat in gas Ab Am (log cfu/g) Phazir 2400 PLS, MPLS RMSECV = 1.063 73
En (log cfu/g) (Polychromix), 1600 RMSECV = 1.198
Lab (log cfu/g) 2400, reflectance RMSECV = 1.256
pH RMSECV = 0.100
Pork meat in gas Bc Am (log cfu/g) RMSECV = 1.079
En (log cfu/g) RMSECV = 1.153
Lab (log cfu/g) RMSECV = 1.296
pH RMSECV = 0.117
Pork meat in gas Ab gas Bc Am (log cfu/g) RMSECV = 0.817
En (log cfu/g) RMSECV = 0.342
Lab (log cfu/g) RMSECV = 0.982
pH RMSECV = 0.039
Duroc pork and Iberian pork M. masseter muscle (correct FieldSpec (ASD), 350 ANN 95 % 74
classification of different 2500, reflectance
breeds of pork)
Rabbit In vivo, live rabbits LabSpec-Pro (ASD), Cluster R2 = 0.316 77
False flax seeds: LD muscle (in ethanol) 3502500, reflectance hierarchical R2 = 0.374
0% (C) Fatty acids (perirenal fat) analysis R2 = 0.86 (M vs. C)
10% (M) R2 = 0.90 (H vs. C)
15% (H) R2 = 0.29 (M vs. H)
Tuna Fat content (%), frozen, Fruit Tester 20 (Fantec), PLS 5.14 , RMSEP , 9.42d 79
glazed tuna 6001000
Fat content (%), frozen, transmittance 3.23 , RMSEP , 5.08d
nonglazed tuna
Fat content (%) , thawed 2.09 , RMSEP , 3.80d
tuna
Fat content (%) (with 3.18 , RMSEP , 4.72d
temperature compensation)
Skipjack fish Fat content (%) Fruit Tester 20 (Fantec), PLS 1.06 , RMSEP , 3.42d 80
6001000, I.N.A.
Meat Fat (%) Phazir 1624 MPLS RMSEP = 4.34 82
Moisture (%) (Polychromix), 1600 RMSEP = 0.63
Protein (%) 2400, reflectance RMSEP = 2.57
Pig Intact carcass fat; correct Phazir 1624 PLS2-DA 60.6 , 93.9 , 96.4 75
classification according to (Polychromix), 1600
feeding regimes (%) 2400, reflectance
a
Am, aerobic mesophiles; ANN, artificial neural networks; cfu, colony-forming units; En, Enterobacteriaceae; INA, information not available; Lab, lactic acid
bacteria; LD, longissimus dorsi; MPLS, multiblock partial least squares; PCA, principal component analysis; PLS, partial least squares; PLS-DA, partial least squares
discriminant analysis.
b
FreshlineTM 3MIX (30:40:30 O2/CO2/N2).
c
FreshlineTM 3MIX 20/5 (5:20:75 O2/CO2/N2).
d
Results vary according to the measured portion and/or wavelength selection.
known to be greatly responsible for harvested fruit and on the loss of quality portable devices are comparable to those
variability in fruit quality. Portable during storage.56,68 Research was also achieved using bench-top spectrome-
NIR spectroscopy could be a solution focused on the plant leaves, not just the ters.52,55,56,59,68 Cayuela and Weiland
to the problem of fruit selection, assess- fruit. There the nutritional status of a compared the performance of two dif-
ing the individual quality characteristics plant was mainly determined as a ferent portable NIR spectrometers and
of the fruits to match them to the function of its concentration of nitro- found that the success of this technique
required internal quality standards. One gen69 and other chemical elements, such is related to the selection of the most
of the main concerns of fruit growers as P, K, Ca, Mg, Fe, Zn, and Mn.70 The suitable instrument for a particular
and distributors is fruit quality after studies showed that NIR spectroscopy is parameter.55 Some of the weaknesses
harvest. The shelf-life and storage con- highly efficient for determining nitrogen of the NIR spectroscopy method include
ditions of various fruit were evaluated to concentration. the need to find a proper calibration
understand the influence of such crop The performance of different types of method and the constant need to keep it
practices as irrigation, fertilization, thin- spectrometers was also tested, determin- updated (especially when the samples
ning, and pruning on the quality of the ing that the results obtained using involved may be subject to periodical or
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Milk (with adulterants: tap water, Fluid milk (% v/v) MicroPHAZIR PLS RMSEP = 4.74 89
whey, synthetic milk, synthetic (Polychromix), 1600
urine, urea, H2O2) 2400, diffuse
reflectance
Milk Fat (%) INA, 6001100, diffuse PLS RMSEP = 0.187 88
Protein (%) reflectance RMSEP = 0.120
DM (%) RMSEP = 0.296
Milk Fat (%) INA, 8001080 PLS 0.066 , RMSECV , 3.62b) 90
Protein (%) DNI, 10501650, 0.078 , RMSECV , 1.87b
Lactose (%) transflectance, RMSECV = 0.081b
Moisture (%) reflectance 0.09 , RMSECV , 0.25b
Temperature (%) 3.01 , RMSECV , 3.30b
Rice wine Alcohol (%) DNI, 11002150, PLS RMSEP = 0.259 86
Nonsugar solids (%) reflectance RMSEP = 0.0134
Tea drinks Tea polyphenols DNI, 11002150, PLS RMSECV = 0.059 85
Free amino acids reflectance RMSECV = 0.005
Beer Bitterness (IBU) PlaScan-SH (OPT MLR R2 = 0.67 83
Total polyphenol (mg/L) Research Inc.), 1200 R2 = 0.64
Total nitrogen (%) 2400, diffuse R2 = 0.69
pH reflectance R2 = 0.64
Beer color (EBC) R2 = 0.34
Apparent extract (%) R2 = 0.53
Alcohol (%) R2 = 0.76
Beer distinction PCA Beers were clustered into three groups
a
DM, dry matter; DNI, development of a new instrument; EBC, European Brewing Convention; H2O2, hydorgen peroxide; IBU, international bitterness unit; INA,
information not available; MLR, multiple linear regression; PCA, principal component analysis; PLS, partial least squares.
b
Results vary according to the type of spectrometer used.
TABLE VI. Reported applications of portable NIR instruments for cereal and feedstock analysis.a
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Wheat leaf Nitrogen content (fresh leaves) FieldSpec (ASD), 350 PLS RMSECV = 0.372 97
(% DM) 2500, reflectance
Nitrogen content (dried leaves) RMSECV = 0.268
(% DM)
Nitrogen content (all leaves) RMSECV = 0.352
(% DM)
LMA (g/m2) RMSECV = 6.300
Maize silage DM (%) INA, 9601700, I.N.A. MPLS RMSECV = 0.97 100
CP (%) RMSECV = 0.12
NDF (%) RMSECV = 0.40
Noodle wheat flours (whole Moisture (%) PlaScan-SH (OPT PLS RMSEP = 0.351 99
noodles) Protein (%) Research Inc.), 1200 RMSEP = 0.526
Ash (%) 2400, diffuse RMSEP = 0.397
Fat (%) reflectance RMSEP = 0.087
Carbohydrate (%) RMSEP = 0.681
Noodle wheat flours Moisture (%) RMSEP = 0.321
(powder) Protein (%) RMSEP = 0.426
Ash (%) RMSEP = 0.402
Fat (%) RMSEP = 0.091
Carbohydrate (%) RMSEP = 0.458
Rice Low-level cadmium-polluted INA, 12002400, diffuse CDA 86.7% 96
Medium-level cadmium-polluted reflectance 83.9%
High-level cadmium-polluted 87.5%
Overall correct classification rate 85.2%
Rice Protein content (%) INA, INA, transmittance PLS RMSEP = 0.279 95
Amylose content (%) RMSEP = 0.52
a
CDA, canonical discriminant analysis; CP, coat protein; DM, dry matter; INA, information not available; LMA, leaf mass per unit area; MPLS, multiblock partial
least squares; NDF, neutral detergent fibers; PLS, partial least squares.
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Soil Total N content (%) DNI, 940, 1050, 1100, BPNN RMSEP = 0.019 91
1200, 1300, and 1550,
reflectance
Soil Soil moisture content (%) DNI, 1450, INA Linear regression RMSEP = 0.85 109
method
Soil Total N content (%) DNI, 940, 1050, 1100, BPNN RMSEP = 0.043 110, 111
1200, 1300, and 1550,
reflectance
Soil Total N DNI, 8002500, PLS R2 = 0.8324 103
Total organic matter reflectance R2 = 0.8833
Soil Organic matter (%) Luminar 5030 PLS RMSEP = 13.353 104
pH (Brimrose), 1100 RMSEP = 1.1126
Ammonium nitrogen 2300, reflectance RMSEP = 23.2595
Nitric nitrogen RMSEP = 106.285
Total kalium RMSEP = 10.1516
Soil Organic C (%) DNI, 16003100, PLS, SMLR 0.18 , RMSEP , 0.27b 105
Moisture content (%) reflectance 1.7 , RMSEP , 1.9b
Clay content (%) 3.6 , RMSEP , 4.4b
Soils Organic C (%) DNI, 16402640, PLS 0.23 , RMSEP , 0.28c 106
Cation exchange capacity reflectance 3.59 , RMSEP , 3.91c
(mEq/100 g)
Moisture content (%) 1.59 , RMSEP , 1.88c
Cattle manure Total C (mg/g) INA, 3601690, INA PCA/PLS RMSEP = 23.4 115
Organic C (mg/g) RMSEP = 0.91
Total N (mg/g) RMSEP = 0.74
C:N (mg/g) RMSEP = 0.87
K (mg/g) RMSEP = 0.83
S (mg/g) RMSEP = 0.73
P (mg/g) RMSEP = 0.61
pH RMSEP = 0.89
Compost (moist) Organic matter (%) AgriSpec (ASD), 350 PLS, PCR 10.1 , RMSECV , 25.7b 116
Compost (oven-dried) 2500, diffuse
reflectance
a
BPNN, back-propagation neural network; DNI, development of a new instrument; INA, information not available; MLR, multiple linear regression; PCA, principal
component analysis; PCR, principal component regression; PLS, partial least squares; SMLR, stepwise multiple linear regression.
b
Results vary according to the calibration method.
c
Results vary according to the data set and moisture content.
is directly related to diet. Therefore, to small size) provided a low-cost solution ucts. Soil organic matter is one of the
meet the requirements of animals and for to the problem of determining wheat and most studied parameters.102108 Some
maximum production, diets must be corn quality in complex environ- experiments using spectroscopy showed
optimized. A study on the performance ments.91,92,94,101 a good correlation between the spectral
of portable NIR spectroscopy instru- Soils and Manures. Studies of soil data and data obtained using standard
ments when applied to the analyses of samples have also used NIR spectros- chemical methods, whereas others sug-
maize silage. The accuracy of the results copy (Table VII). Soil samples for these gested the need for further verification of
was very close to that obtained using a studies were usually collected at the results over a wider range of soil
laboratory NIR instrument.100 Once surface or at known depths. In those samples. Gomez et al. exposed for the
again, a comparison was made of situations, samples were first sieved and/ first time the potential of satellite hyper-
various regression techniques used dur- or dried (either oven- or air-dried) and spectral data for the prediction of the
ing the cereal analysis, and the results then analyzed. Different types of soils organic carbon content of soil. They
favored the artificial neural networks were investigated and compared. Some demonstrated good accuracy of the
(ANN) technique.93 Studies of the reports have been made of researchers model no matter what number or range
development of new portable NIR attempting in-field soil analyses using of soil samples used.108 Soil moisture
spectroscopy devices were also done specific contact probes. The measure- was also a commonly studied parame-
on cereals. It was concluded that the ment and control of soil characteristics ter,105,106,109 and one of these studies
portable equipment (which had a simple are particularly important for the proper used spectroscopy to generate spatial
structure, low-power consumption, and development of the agricultural prod- distribution maps.109 Total nitrogen
Spectrometer, Multivariate
spectral range (nm), analytical
Species Attributes acquisition mode methods Results Reference
Grapes (fresh berries) SSC (8Bx) AvaSpec-2048 (Avantes), PLS RMSEP = 1.48 117
TA (g tartaric acid/dm3) 4001000, reflectance RMSEP = 1.48
pH RMSEP = 0.15
PA (mg/dm3) RMSEP = 133.90
EA (mg/dm3) RMSEP = 81.87
Grapes (homogenized samples) SSC (8Bx) RMSEP = 0.95
TA (g tartaric acid/dm3) RMSEP = 1.12
pH RMSEP = 0.13
PA (mg/dm3) RMSEP = 129.00
EA (mg/dm3) RMSEP = 77.70
Grapes (fresh berries) SSC (8Bx) PLS-DA 77.1%
TA (g tartaric acid/dm3) 68.6%
Grapes (homogenized samples) SSC (8Bx) 88.7%
TA (g tartaric acid/dm3) 83.1%
Grapevine leaf (Syrah, Merlot, Water potential Luminar 5030 PLS RMSECV = 1.48 126
Cabernet Sauvignon) (Brimrose), 11002300
Grapevine leaf (Cabernet transmittance RMSECV = 1.08
Sauvignon)
Grapes (Cabernet Sauvignon) SSC (8Bx) INA, 5901090, INA PLS 0.61 , RMSECV , 1.15b 118
Grapes (Cabernet Sauvignon, TA (g tartaric acid/dm3) INA, 4001000, PLS 0.2 , RMSEP , 0.4b 119
Sangiovese, Merlot) Maturity index (sugar/acidity reflectance 2.8 , RMSEP , 6.0b
ratio)
PA (mg/L) 101.3 , RMSEP , 125.7b
pH 0.0 , RMSEP , 0.1b
Reductant sugars (g/L) 6.6 , RMSEP , 12.2b
Grapes (Cabernet Sauvignon, SSC (8Bx) USB 2000 (Ocean PLS 1.010 , RMSEP , 1.267c 120
Carmene`re, Merlot, Pinot pH Optics), 6401300, 0.088 , RMSEP , 0.159c
Noir, Chardonnay) Anthocyanin concentration INA 0.183 , RMSEP , 0.315c
(mg/g)
Grapes SSC (8Bx) Luminar (Brimrose), PLS RMSECV = 1.9 121
pH 11002300, RMSECV = 0.16
TA (g/L) transmittance RMSECV = 1.89
Anthocyanin concentration RMSECV = 0.11
(mg/g)
Grapes SSC (8Bx) INA, 4001100, MLR RMSECV = 1.12 127
Acidity (g H2SO4/L) interactance RMSECV = 1.35
Water content (%) RMSECV = 1.89
Olives (Moraiolo, Dolce di Total phenols Luminar 5030 PLS 0.451 , RMSECV , 1.672c 128
Andria, Nocellara Etnea) Verbascoside (Brimrose), 1100 0.074 , RMSECV , 0.210c
Oleuropein 2300, transmittance 0.175 , RMSECV , 0.365c
3,4 DHPEA-EDA 0.237 , RMSECV , 1.138c
Olives (Arbequina, Picual, Oil content (%) Luminar 5030 PLS 1.52 , RMSECV , 1.99c 129
Verdial de Huevar, Manzanilla Moisture (%) (Brimrose), 1100 1.65 , RMSECV , 2.26c
de Sevilla) 2300, transmittance
Olives Fruit moisture (%) LabSpec (ASD), 350 PLS RMSECV = 1.741 130
DM (%) 2500, reflectance RMSECV = 1.710
Oil content (fresh weight) RMSECV = 0.901
(%)
Oil content (DM) (%) RMSECV = 1.619
Fruit maturity index (%) RMSECV = 0.465
Oil-free acidity (%) RMSECV = 0.030
a
DM, dry matter; 3,4 DHPEA-EDA, 3,4-dihydroxyphenolethanol-elenolic acid di-aldehyde; EA, extractable anthocyanins; H2SO4, sulfuric acid; INA, information
not available; MLR, multiple linear regression; PA, potential anthocyanins; PLS, partial least squares; PLS-DA, partial least squares discriminant analysis; SSC,
soluble solids content; TA, titratable acidity.
b
Results vary according to the calibration method.
c
Results vary according to the species.
content,103,104,110,111 pH,104 soil color spectra and laboratory-collected spectra al. simultaneously investigated the spec-
and mineral composition,112 and clay (after drying the samples). Surprisingly, tral properties of soil and wheat, with the
content have also been predicted.105,112 and contrary to what was expected, the aim of discriminating between conven-
Viscarra Rossel et al. compared models field-collected spectra provided better tional and conservation agricultural
for clay content using field-collected predictions of clay content.112 Hache et practices. The soil spectra were useful