Sei sulla pagina 1di 50

This article was downloaded by: [CSIRO Library Services]

On: 17 February 2014, At: 04:00


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Applied Spectroscopy Reviews


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/laps20

The Performance of Visible, Near-, and


Mid-Infrared Reflectance Spectroscopy
for Prediction of Soil Physical, Chemical,
and Biological Properties
ab a ac
José M. Soriano-Disla , Les J. Janik , Raphael A. Viscarra Rossel ,
a ad
Lynne M. MacDonald & Michael J. McLaughlin
a
CSIRO Sustainable Agriculture Flagship Program, CSIRO Land and
Water, Glen Osmond, South Australia, Australia
b
Department of Agrochemistry and Environment, University Miguel
Hernández of Elche, Elche, Spain
c
CSIRO Land and Water, Bruce E. Butler Laboratory, Canberra,
Australia
d
Soil Science, School of Agriculture, Food and Wine, Waite Research
Institute, University of Adelaide, Glen Osmond, South Australia,
Australia
Accepted author version posted online: 01 Jul 2013.Published
online: 12 Aug 2013.

To cite this article: José M. Soriano-Disla, Les J. Janik, Raphael A. Viscarra Rossel, Lynne M.
MacDonald & Michael J. McLaughlin (2014) The Performance of Visible, Near-, and Mid-Infrared
Reflectance Spectroscopy for Prediction of Soil Physical, Chemical, and Biological Properties, Applied
Spectroscopy Reviews, 49:2, 139-186, DOI: 10.1080/05704928.2013.811081

To link to this article: http://dx.doi.org/10.1080/05704928.2013.811081

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014
Applied Spectroscopy Reviews, 49:139–186, 2014
Copyright © 2014 Crown copyright
ISSN: 0570-4928 print / 1520-569X online
DOI: 10.1080/05704928.2013.811081

The Performance of Visible, Near-, and Mid-Infrared


Reflectance Spectroscopy for Prediction of Soil
Physical, Chemical, and Biological Properties

JOSÉ M. SORIANO-DISLA,1,2 LES J. JANIK,1


RAPHAEL A. VISCARRA ROSSEL,1,3
LYNNE M. MACDONALD,1
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

AND MICHAEL J. McLAUGHLIN1,4


1
CSIRO Sustainable Agriculture Flagship Program, CSIRO Land and Water,
Glen Osmond, South Australia, Australia
2
Department of Agrochemistry and Environment, University Miguel Hernández
of Elche, Elche, Spain
3
CSIRO Land and Water, Bruce E. Butler Laboratory, Canberra, Australia
4
Soil Science, School of Agriculture, Food and Wine, Waite Research Institute,
University of Adelaide, Glen Osmond, South Australia, Australia

Abstract: This review addresses the applicability of visible (Vis), near-infrared (NIR),
and mid-infrared (MIR) reflectance spectroscopy for the prediction of soil properties.
We address (1) the properties that can be predicted and the accuracy of the predic-
tions, (2) the most suitable spectral regions for specific soil properties, (3) the number
of predictions reported for each property, and (4) in-field versus laboratory spectral
techniques.
We found the following properties to be successfully predicted: soil water content,
texture, soil carbon (C), cation exchange capacity, calcium and magnesium (exchange-
able), total nitrogen (N), pH, concentration of metals/metalloids, microbial size, and
activity. Generally, MIR produced better predictions than Vis-NIR, but Vis-NIR out-
performed MIR for a number of properties (e.g., biological). An advantage of Vis-NIR
is instrument portability although a new range of MIR portable devices is becoming
available. In-field predictions for clay, water, total organic C, extractable phosphorus,
total C and N appear similar to laboratory methods, but there are issues regarding, for
example, sample heterogeneity, moisture content, and surface roughness.
The nature of the variable being predicted, the quality and consistency of the
reference laboratory methods, and the adequate representation of unknowns by the
calibration set must be considered when predicting soil properties using reflectance
spectroscopy.

Keywords: Mid-infrared, near infrared, prediction, soil, spectroscopy, visible

Address correspondence to Jose M. Soriano-Disla, CSIRO Sustainable Agriculture Flagship


Program, CSIRO Land and Water, PMB 2, Glen Osmond, South Australia 5064, Australia. E-mail:
jose.sorianodisla@csiro.au

139
140 J. M. Soriano-Disla et al.

Introduction
Soil analysis using traditional laboratory technology (e.g., wet chemistry) can be pro-
hibitively expensive and often time consuming, prompting the use of pedotransfer functions
as a substitute (1, 2). As a consequence, these expensive soil analyses are often restricted to
a few samples or to samples that are bulked from throughout an area to provide represen-
tative composites. Such data will have little or no information on the spatial variability of
soil. Development of rapid and more cost-effective methodologies is required to meet this
demand and surrogate methods based on spectroscopic data in the ultraviolet (UV), visible
(Vis), near-infrared (NIR), and mid-infrared (MIR) frequency regions (or combinations)
have good potential (3, 4).
Previous reviews have largely focused on the use of Vis-NIR for soil analysis (5–9),
MIR (3) and UV-Vis-NIR, Vis-NIR, NIR, and MIR (10, 11). Specific reviews have been
published for soil carbon (C) predicted by infrared spectroscopy (12, 13), the evaluation of
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

soil fertility using various MIR techniques (14), and laboratory and proximal Vis-NIR and
MIR sensing predictions (4, 15).
Despite the extensive literature directed toward the prediction of soil properties from
UV, Vis, NIR, and MIR spectral data, more research is needed to address a number of crucial
questions that still remain to be answered. The answers to these questions are essential for
a balanced assessment and development of cost-effective and reliable spectroscopy-based
soil analysis:
1. What can be predicted and what is the achievable performance of these predictions?
2. Which spectral regions are the most suitable for these predictions?
3. How many examples of predictions have been reported for each property for the
different Vis-infrared spectral ranges?
4. What is the performance of in-field versus laboratory (benchtop) calibrations?
The main objective of this review is to attempt to answer these questions.
Here, we provide a comprehensive review on the prediction accuracies reported for soil
physical, chemical, and biological properties using the MIR, NIR, and Vis-NIR. Consider-
ably fewer studies using single UV, Vis, and combined spectral ranges are also evaluated
(UV-Vis-NIR, NIR-MIR and Vis-NIR-MIR). Based on the findings of the review, we make
recommendations for the use of reflectance spectroscopy as a quantitative tool for soil
science.

Visible-Near and Mid-Infrared Diffusive Reflectance Spectroscopy


There is an increasing interest in using diffusive reflectance infrared spectroscopy for the
rapid and cost-effective prediction of soil physical, chemical, and biological properties.
Spectral reflectance methods are nondestructive, suitable for neat (undiluted) soils, and
provide spectra that are highly characteristic of the soil type and composition, thus allowing
for the analysis of many soil properties (3, 10, 16, 17). Furthermore, prediction methods
based on diffuse reflectance spectroscopy require minimal or no sample preparation, avoid
the use of environmentally harmful extractants in the laboratory (3, 17), and are easily
adaptable for proximal sensing, especially with Vis-NIR (10, 15), further expanding the
uses of reflectance spectroscopy in recent years.
Infrared spectra of soils contain extensive information on the molecular and composi-
tional chemistry that can be used for a better understanding of soils as a complete environ-
mental system and as a long-term resource (10, 18). Thus, Vis and infrared spectroscopy
Soil Properties 141

have been also used for a more general assessment of soil organic matter (SOM) composi-
tion, soil quality (capacity to function), mapping, classification, and fertility (19–28).
Infrared spectroscopy relies on the absorbance of radiation at molecular vibrational fre-
quencies. These frequencies can occur for relatively light vibrating C H, N H, and O H
groups containing hydrogen, as well groups of “heavier” atoms C O, C N, N O, C C
in organic materials as well as Al-O, Fe-O, and Si-O in minerals (3, 14, 29, 30). Overtone
and combination vibrations of the relatively light atoms involving hydrogen dominate the
NIR (700–2,500 nm; 4,000–14,286 cm−1), while these, plus bonds from the heavier atomic
groups, absorb in the MIR (2,500–25,000 nm; 400–4000 cm−1). Electronic transitions
absorb in the ultraviolet (250–400 nm; 25,000–40,000 cm−1) and visible (400–700 nm;
14,286–25,000 cm−1) regions (10, 30). The infrared regions specific to certain analytes
are described in both wavelength (nm) and wavenumber (cm−1) units, commonly used by
the Vis-NIR and MIR community, respectively (showing first the regions with units as
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

originally published).
The main frequencies and wavelengths of spectral absorption in the MIR and Vis-
NIR for major soil components can be found in Table 1, comprising quartz, clay minerals
(kaolinite, smectite/illite), carbonates, aluminum (Al)/iron (Fe) oxyhydroxides, Fe oxides,
and SOM. Band assignments for the Vis-NIR for soil organic carbon, pH, and clay content
can be found in Viscarra Rossel and Behrens (31). Visible absorption in soil spectra is
primarily associated with minerals, mostly containing Fe oxides (e.g., hematite, goethite)
(32). Soil organic matter (which is typically dark) can also have broad absorption in the Vis
region, but this appears often as an upward shift in the reflectance (8).
Near-infrared spectroscopy has some advantages in terms of cost and portability of the
instruments, and samples can be scanned in water and glass containers. However, the NIR
region is insensitive to quartz (instruments often use quartz optics) and NIR region overtones
and combination bands can overlap, making the spectra more difficult to characterize than
in the MIR (10, 13). In contrast to NIR, peaks in the MIR are frequently better resolved
and much more intense, and more spectral information is available for MIR (3, 33, 34).
This can be also an issue in MIR. For example, Fe oxide minerals have frequencies that
overlap with other soil mineral peaks near 600–700 cm−1 (14,286–16,667 nm), making
them difficult to identify. The COO− anion and water frequencies have been found to
overlap with aromatic groups near 1600–1570 cm−1 (6,250–6,369 nm) (35). Mid-infrared
spectra of SOM are complicated by the presence of many inorganic MIR absorbing species,
including Si-O groups in quartz and C O groups in carbonates (16).
Prediction of soil attributes in unknown soils can be achieved by deriving models from
the reference laboratory data and corresponding spectra. Some attributes (e.g., carbonates,
SOM, and Fe oxides) can be predicted directly based on the presence of characteristic
chemical bonds or molecules that absorb at specific wavelengths. Alternatively, calibrations
might be based on the correlations between the attribute of interest and Vis-infrared–active
compounds. These correlations allow for the development of calibrations—for example,
some metallic elements that are not Vis-infrared active or where the peaks are too weak to
be detected or overlap with peaks from the soil matrix.

Calibration Considerations
In assessing the reported predictions of soil properties from Vis-NIR and MIR spectra,
we first need to discuss issues relevant to the development, accuracy, or robustness of the
underlying regression models.
142 J. M. Soriano-Disla et al.

Table 1
Approximate wavenumbers and wavelengths of spectral absorptions in the MIR and Vis-
NIR regions for major soil components

Soil Wavenumbers Wavelengths


component (cm−1) (nm) References
MIR Quartz (sand) 1100–1000 9,091–10,000 (16, 29)
Clay minerals 3690–3620 2,710–2,762 (16, 29, 31)
Kaolinite
3620–3630 and 2,755–2,762 and (16, 29, 31)
3400–3300 2,941–3,030
Smectite and
illite
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Carbonates 1430 and 2520 6,993 and 3,968 (16, 29, 31)
Iron oxides 600–700 14,286–16,667 (29)
Iron 3100, 900, 800 3,226, 11,111, (29)
oxyhydroxides 12,500
Organic matter 2930–2850 Alkyl 3,413–3,509 (16, 29, 31)
(–CH2 )
1670 and 1530 5,952 and 6,535 (16, 29, 31)
Protein amide
(OC-NH)
1720 Carboxylic 5,814 (16, 29, 31)
acid (COOH)
1630 Water 6,135 (16, 29, 31)
associated
1600 and 1400 6,250 and 7,143 (16, 29)
Carboxylate
anion ( COO−)
1600–1570 6,250–6,369 (29)
Aromatic
groups
Vis-NIR Water 7143 and 5263 1,400 and 1,900 (31, 104, 194)
Clay minerals 7143 and 4545 1,400 and 2,200 (31)
Kaolinite
4545, 4274, and 2,200, 2,340, and (31)
4090 Illite 2,445
4545 Smectite 2,200 (31, 104)
Carbonates 4283 2,335 (31, 104)
Iron oxides 25,000, 22,222, 400, 450, 500, (31)
20,000, 15,385, 650, and 900
and 11,111
Organic matter 9091, 6250, 5882, 1,100, 1,600, (31, 104, 194, 195)
5556, 5000, and 1,700, 1,800,
4167–4545 2,000, and
2,200–2,400
Soil Properties 143

Data Size and Coverage


The selection of representative calibration and internal validation samples (to derive and
train the model, respectively) and an independent validation or “test” set (to check the
predictive performance of the model) are important requirements when developing models
for prediction of properties using spectroscopy techniques. The calibration samples should
cover the variability expected in the full sample set and future unknowns (spectra and
reference analytical data) (36, 37), and the validation (test) set must be independent of the
calibration set in order to avoid an optimistic assessment of predictive performance. The
number of samples available for calibration and validation may be influenced largely by
the cost and complexity of acquiring the reference data (e.g., pH versus C pool chemistry).
The first step in model development is to set aside a randomly selected test set (normally
25% of the full data set) (38). The next step is to derive, train, and optimize the model
by either splitting into calibration and internal validation set, as discussed above, or using
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

cross-validation. The first approach is commonly used in data-rich situations, with, as a


rule of thumb, two thirds and one third of the data allocated to calibration and internal
validation sets, respectively (38).
Cross-validation is a special case of calibration training, particularly suited for use
where few data are available in an attempt to utilize as much of the value of the sample
reference data as possible. It is commonly used to validate the calibration during model
training and tuning the model (assessing performance and optimum number of factors),
without the use of a separate internal validation set. With this method, each sample (leave-
one-out cross-validation), or group of samples, is removed in turn from the calibration set
and its value is predicted as an unknown from models built from the remaining samples in
the set. Cross-validation generally gives an optimistic assessment of the actual performance
of the models, particularly when the number of samples is small (36). For this reason, it is
better to assess the performance of the model with an independent test set.
Large spectral libraries, ideally accounting for all of the expected variability in the
spectra of the soil types, can be used to build calibrations for the prediction of soil prop-
erties in unknowns. Ideally, the library spectra should match those of future unknowns.
Occasionally, however, there are the spectra of unknowns that do not match those from the
calibration spectra (outliers). This can occur when the calibration samples are too restricted,
as in site-specific applications, or the spectral library is too small. Shepherd and Walsh (39)
and Viscarra Rossel et al. (40) suggested how to build spectral libraries and how to deal
with unknowns.
An alternative to the use of a single large calibration is the development of site-specific
calibrations. Site-specific models are best suited to relatively homogeneous areas or to
smaller-scale applications. Large models usually represent a wide range of sample char-
acteristics, often from varying geological, geographic, or climatic regions (41). According
to Reeves (13), the question of whether calibrations for soil C beyond a field scale can be
useful is directly related to what basis (e.g., analyte levels, texture, soil type or combina-
tions thereof) soils can be selected for or included in a single calibration. Udelhoven et al.
(42) and Nduwamungu et al. (7) reported that the development of reliable NIR models for
soil chemical properties should be limited to geologically homogeneous areas, where “un-
known” samples are generally adequately represented within the group of samples used for
calibration. Alternatively, calibrations from vast data sets can be developed, thus assuming
that any unknown samples would be spectrally similar to those in the calibration set. Such
spectral libraries have been built for the Vis-NIR for Africa (39), Australia (43), and an even
larger worldwide set (4). Shepherd and Walsh (39) found that inclusion of local samples
144 J. M. Soriano-Disla et al.

in calibrations, or using local calibrations, provided more robust models for the prediction
of a number of soil properties. An alternative strategy for achieving a local calibration
would be to implement some kind of calibration selection (i.e., a targeted calibration based
on the spectral characteristics of the unknown sample) (44, 45). In some cases, it may be
advantageous to split the calibration set simply into a small number of separate calibrations
based on clustering or classification criteria and then use these specific calibrations for the
particular group or class that the unknowns fall into. Viscarra Rossel and Webster (43) used
regression trees for dividing large and diverse data sets to provide local models for the
prediction of soil properties.
There are formal statistical techniques available for selecting the spectra for an optimum
calibration set. The optimization can involve the selection of a subset of spectra that
represent a full spectral library and also the removal of spectra that can have an adverse effect
on the calibration (e.g., outliers). The selection of a subset of spectra is commonly based on
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

the measurement of the similarity between the spectra of the samples (46). Ramirez-Lopez
et al. (47) have shown that spectral similarity can reflect the soil composition similarities
by using various distance metric algorithms.
Soil properties whose concentration is highly variable (e.g., soil available nutrient
concentrations) require a high sampling effort because they vary even on a small scale
throughout a cultivated field. Such high-density sampling is expensive, especially if it
includes reference laboratory analysis. Optimum calibrations for predictive purposes must
be regularly updated with the addition of new samples in relation to soil variability (39–41,
43). Ideally, and for proximal soil sensing, calibration sampling must cover the property
and geographical space as well as boundaries and other transition zones (4).

Accuracy and Determination of Reference Data


A major problem in predicting soil properties from spectra is the use of inappropriate,
incorrect, or mismatched reference laboratory data. Firstly, there are obvious assumptions
in the use of reliable standard analytical methods for obtaining reliable and meaningful
calibrations. In many cases, the problem comes when combining soils from different data
sets with the attribute of interest determined by different analytical methods that do not
give the same result (48). This issue is particularly relevant for total organic C (TOC),
where the Walkley and Black and the dry ignition (e.g., Leco) methods give quite different
results (49). Some other examples are charcoal-C (50) and soil pH (51). In other cases, the
method itself, even when all of the soils have been analyzed with the same method, might
introduce confounding factors that negatively affect the models. This is, for example, the
case of particle size analysis, where the effect of not removing carbonates from calcareous
soils can have a major impact on the calibrations (52).

Sample Preparation (Drying, Sieving, and Grinding)


Given the heterogeneous nature of soils, some sample preparation is often required to
optimize reproducibility and accuracy of spectroscopic predictions. Sample preparation is
usually more important for the MIR compared to the Vis-NIR (53). This is partly due to the
adverse effects of sample heterogeneity on the relatively smaller sampling area of the MIR
beam (∼2–3 mm diameter compared to >5 cm, as often found in the Vis-NIR), requiring
either repeated sample scanning or alternatively soil sieving to <2 mm nominal particle
diameter. Additional grinding of soil to <100 μm can reduce the inter- and intraparticle
variability. Grinding of soil aggregates can have a substantial effect on spectral absorbance.
Soil Properties 145

In the NIR, the overall reflectance increases and peaks may reduce in intensity. The effect is
particularly strong for soils high in clay (8). There is also distortion of MIR spectra due to
specular reflectance. The use of different particle sizes may thus result in variable accuracy
of infrared predictions for similar soil properties (7), and great care needs to be taken to
ensure that the same soil preparation protocols are strictly maintained for the most accurate
predictions.
For laboratory-based application, and due to small beam size and intra and interparticle
heterogeneity, optimum (i.e., better representing the soil sample) MIR spectra are then
generally obtained when samples are milled to <100 μm and dried at 40◦ C (54). In
field conditions, the presence of water can adversely affect MIR spectra due to significant
surface reflection from water films. In the NIR, peaks around 1,400 and 1,900 nm (7,143
and 5,263 cm−1), common in clay soils, are strongly affected by soil water (55). As for the
MIR, water can act as a reflective film for Vis-NIR in samples with water contents greater
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

than the field capacity. Although dependent on the soil property, the necessity of analyzing
field-moist soils (at water contents less than field capacity) is not a major impediment to
the deployment of commercial NIR soil analysis technology (55–57). However, and for
some properties such as TOC, the moisture content can negatively affect the accuracy of
the predictions (58).

Instrumentation and Infrared Spectroscopy Techniques (Technical Characteristics)


Predictions for the same soil properties, under standard conditions, can be different ac-
cording to the instrumentation used (13). This may be due to the different specifications
of the instrumentation used: resolution, sample accessories, instrument stability, and en-
ergy intensity for the different Vis and infrared regions. A specific description of different
spectroscopic instruments can be found in Reeves (13).
Ultraviolet-visible instruments are usually configured specific for this spectral range,
but in some cases they are combined with NIR and MIR. Instruments come in a number
of configurations, using either a cuvette or an integration diffuse reflecting sphere, and in a
number of options; for example, Fourier-transform (FT) instruments, scanning monochro-
mators, filter based, and diode array. The variety of instruments used presents problems
with respect to data compatibility and calibration transfer (13). In the case of MIR spec-
trometers, these are equipped with standardized diffuse reflectance accessories and they
can deliver almost identical spectra irrespective of instrument manufacture.
Low-cost portable instruments for the Vis-NIR have been widely used. These in-
struments often have wavelength range and noise level limitations compared to benchtop
instruments. Mid-infrared portable instruments are also available but, as discussed, potential
problems related to water content and sample heterogeneity exist.
Among different infrared spectroscopy techniques for recording spectra, diffuse re-
flectance infrared Fourier transform (DRIFT) is commonly used for soil analysis (15).
When the radiation is focused onto a soil sample, part of the incoming radiation is reflected
back in all directions by the soil particles. This reflected radiation is detected by the spec-
trometer optics and subsequently used as the spectral inputs for regression analysis. An
alternative to DRIFT is attenuated total (internal) reflectance (ATR) or evanescent wave
spectroscopy, which utilizes the absorption of radiation from the interface between a high
refractive index (e.g., germanium, zinc selenide, or metal halides), reflecting crystal or fiber
and the sample surface. This method relies on intimate optical contact between the sample
and reflecting surface and is only effective to a few wavelengths of the infrared radiation
(59, 60). Another available technique for infrared spectral acquisition is photoacoustic
146 J. M. Soriano-Disla et al.

(PAS) spectroscopy (61, 62). This technique is based on the absorption of infrared radiation
by the analyte molecules and detection of the modulated thermal energy, absorbed within
a sealed layer of gas (e.g., dry air or helium) surrounding the sample, as sound with a
sensitive microphone. Though very effective for some applications, PAS measurements
require precise setup and scan times are relatively long (sometimes up to 15–30 min for
weakly absorbing samples) but provide enhanced surface characterization (61, 62).

Spectral Overlap, Interpretation, and Reflection


Peaks and band profiles related to the attribute of interest can overlap with those of other soil
compounds, thus hiding the details contained in the band profiles. These overlapped spectral
peaks can often be “resolved” by sharpening the peaks using derivative preprocessing (63),
but interpretation of these new sharper peaks may still present difficulties.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Peaks may appear as positive intensities or negative as a “valley” or trough between


two successive peaks. Unfortunately, there is no substitute to experience in order to avoid
wrong assignment of peaks, which can lead to erroneous conclusions when interpreting
spectral information. For example, numerous quartz peaks in the MIR below 2,000 cm−1
(5,000 nm) almost completely overlap with those of SOM, but they are in some instances
interpreted as “organic matter” peaks (64, 65).
A very strong optical aberration or peak distortion (called restrahlen) can occur as the
result of the contribution of refractive index to the peak shape resulting from mirror-like
reflection from the surfaces of large soil particles. This can distort or even completely
obliterate strong absorbance peaks for the fundamental vibrations of carbonates, quartz,
and other clay minerals in the MIR, resulting in inverted or derivative shaped peaks that can
lead to difficulties in identifying the contributions of major soil minerals to MIR spectra,
particularly at frequencies less than 2000 cm−1 (16, 66). Despite these difficulties with band
distortion, the strong reflectance effect can be used to enable useful information on effects of
particle sizes to be gleaned from the spectra. Furthermore, in other cases where significant
particle heterogeneity occurs, the incident radiation only impinges on the outermost layers
of the soil particles, thus hiding the internal structure from analysis. This can be a major
issue where fine clay or Fe oxides may be coating large quartz sand grains (67), but its
effect can be reduced by homogenizing the particles with grinding.

Spectroscopic Modeling
In order to model the complex relationship between spectral signatures and a soil property,
multivariate regression methods have an advantage over simple bivariate relationships
based on, for example, peak intensity measurements. A range of multivariate techniques
was compared by Viscarra Rossel and Behrens (31) and will not be discussed in detail in this
review: multiple linear regression; partial least squares stepwise linear regression (PLSR);
principal components regression; multivariate adaptive regression splines; support vector
machines; random forest; boosted trees; and the use of wavelet analyses and artificial neural
networks. Of these, PLSR has been perhaps the most popular for soil quantitative analysis
(3, 10, 17) in view of the easy access to software and relatively easy understanding of the
principles behind the relationships between soil properties and the spectral contributions
to the regression models. Partial least squares models are easy to derive and interpret, are
rapid, and are insensitive to colinearity (68). However, other multivariate methods (e.g.,
that account for nonlinear responses such as support vector machines and artificial neural
networks) may perform better (31, 43, 69–72).
Soil Properties 147

Model Performance
The performance (accuracy and precision) of a spectroscopic model is generally assessed
in terms of its coefficient of determination (R2) and root mean square error (RMSE). The
predicted results also have errors, invariably larger than the calibration, and performance of
the predictive capabilities of the model, based on a validation or test set, can be expressed
in terms of the Pearson correlation coefficient (r), RMSE, and the ratio of SD of the
reference soil property values to RMSE (RPD) (73). The RPD allows comparison of
model performance across different data sets. There are also other ways of describing the
residuals in terms of, for example, predicted residual error sum of squares, bias (mean
error) or imprecision (SD of error) (see 38, 43). The inaccuracy (RMSE) embraces both
the bias and the imprecision, and for a complete understanding of the error these three
statistics need to be provided (43). As discussed by Reeves and Smith (74) and Viscarra
Rossel and Webster (43), the values of RPD should be evaluated in conjunction with other
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

assessment statistics previously provided in order to assess the performance of predictive


infrared models.
Unfortunately, the RPD and error estimators are not available in all published articles
(errors are frequently not comparable due to different ranges of data or for log/root trans-
formed variables), whereas R2 is common to all reported studies and has been used here as
a crude measure of the precision of predictions (11).
In the case of models using latent variables (e.g., in PLSR, principal components
regression), the performance criteria must additionally consider the number of factors
used in the model (38)—not too large and not too small, reducing the risk of over- and
underfitting, respectively.

Calibration Transferability
Calibrations obtained in one instrument are usually unsuitable for use in a second instru-
ment due to dissimilarities between spectral point spacing and range, abscissa shift, and
ordinate response. However, calibration transfer can be carried with a number of alternate
strategies to allow calibrations derived on one instrument (“master”) in a second (“slave”)
instrument by conversion of the embedded spectral data into formats compatible with the
new instrument. Infrared spectra can be adjusted with methods such as piecewise direct
standardization and the patented method of Shenk and Westerhaus (75), which use multiple
spectra from two instruments to be matched (76, 77).

Review of the Prediction of Soil Properties


Following the above discussion on the relevant issues for the development of spectroscopic
models, we now review what has been reported in the literature for a range of soil properties,
the specific spectral ranges used, and the statistical performance of these predictions. A large
number of soil properties have been predicted using multivariate chemometric regression
modeling derived from reference soil data and UV, Vis, NIR, and MIR spectra (Table 2).
The main focus of the literature has been on soil chemical properties.
The median R2 values of validation results for prediction of soil properties for the
different single and combined spectral ranges are shown in Table 3. When assessing this
summary information it is important to consider the calibration issues discussed above, be-
cause the literature demonstrates a diverse range of approaches that influence the reported
prediction performance indicators. It should be appreciated that conclusions based on the
148 J. M. Soriano-Disla et al.

Table 2
List of soil variables predicted by reflectance spectroscopy together with chemometric
techniques as reported in the literature

Physical properties
• Aggregation
Aggregation (2, 1, 0.50, 0.25 mm), aggregation (non-erode), aggregation (stability),
macroaggregation, mean geometric diameter, mean weight diameter.
• Soil water
Moisture, field capacity, permeability wilting point, water repellence.
• Particle size
Bulk density, clay, sand, silt, surface area.

Chemical properties
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

• Carbon and organic matter


C (alkyl), C (alkyl)/C (aromatic), C (alkyl)/C (O-alkyl), C (aromatic), C
(aromatic-organic), C (carbonyl), C (charcoal), C (clay fraction), C (dissolved hot
water), C (dissolved-organic), C (inert), C (inorganic), C (labile fractions), C (mineral
associated), C (O-alkyl), C (organic), C (organic-particulate), C (organic)/N (tot), C
(organic)/S (tot), C (organic)/silt clay, C (particulate organic matter), C (resistant
organic C), C (sand fraction), C (sand + stable aggregates), C (silt fraction), C (silt +
clay without resistant organic C), C (stable fractions), C (total), C13, C 375◦ C, C/N,
C/N (clay fraction), C/N (clay fraction), C/N (sand fraction), C/N (silt fraction), C/N
(particulate organic matter), C/P, fulvic acids, fulvic acids, GC-FAME equivalent
chain length 16 peak area, humic acids, humic acids, humus, loss of ignition (LOI),
organic matter, organic matter particulate, organic matter/LOI, phenolic oxidation
products.
• Cation exchange capacity
Cation exchange capacity (CEC), CEC/clay, exchangeable cations, base saturation.
• Macronutrients
Ca (exchangeable), Ca (Mehlich I), Ca (Mehlich III), Ca (oxalate), Ca (acid digestion),
Ca (X-ray fluorescence, XRF), Ca/Mg, K (exchangeable), K (AL extraction), K
(ammonium acetate), K (ammonium acetate and acetic F), K (ammonium lactate), K
(NH4 OAc), K (bicarbonate), K (calcium-acetate-lactate), K (Mehlich I), K (Mehlich
III), K (acid digestion), K (XRF), Mg (exchangeable), Mg (AL extract), Mg
(ammonium lactate), Mg (CaCl2 ), Mg (calcium-acetate-lactate), Mg (Mehlich I), Mg
(Mehlich III), Mg (oxalate), Mg (acid digestion), Mg (XRF), N (amino sugar), N (clay
fraction), N (H2 SO4 ), N (hot water), N (KCl), N (particulate organic matter), N
(particulate organic matter N in 212-μm fractions), N (particulate OM N in 53-μm
fractions), N (particulate organic matter N in fractions <53 μm), N (sand fraction), N
(silt fraction), N15, NH4 +-N, NO3 −-N, N (mineral), N (organic), N (tot), N supply, P
(AL extract), P (ammonium acetate and acetic F), P (ammonium fluoride), P
(ammonium lactate), P (bicarbonate), P (Bray I), P (Bray II), P (Burriel-Hernando), P
(calcium-acetate-lactate), P (Colwell), P (H2 O), P (HCl-NH4 F), P (K2 SO4 ), P
(Mehlich I), P (Mehlich III), P (NaHCO3 ), P (Olsen), P (organic), P (oxalate), P
(vanadomolybdophosphoric), P (acid digestion), P (combustion + HCl), P (XRF), P
(absorptive coefficient), P (buffer), P (capacity), P (resins), P (sorption), S (acid
digestion), S (dry combustion), S (XRF), S (ammonium acetate and acetic F).
(Continued on next page)
Soil Properties 149

Table 2
List of soil variables predicted by reflectance spectroscopy together with chemometric
techniques as reported in the literature (Continued)
• Micronutrients
B (CaCl2 ), Cl, Co (acid digestion), Co (XRF), Cu (DTPA), Cu (Mehlich III), Cu (acid
digestion), Cu (XRF), Fe (citrate bicarbonate dithionite), Fe (dithionite), Fe
(dithionite-citrate), Fe (DTPA), Fe (Mehlich I), Fe (Mehlich III), Fe (oxalate), Fe oxides
(Na dithionite-citrate), Fe (water/H2 SO4 ), Fe oxides (DCB), Fe (acid digestion), Fe
(XRF), Mn (CaCl2 ), Mn (exchangeable), Mn (DTPA), Mn (Mehlich I), Mn (Mehlich
III), Mn (dithionite), Mn (oxalate), Mn (water/H2 SO4 ), Mn (acid digestion), Mn (XRF),
Na (adsorption ratio), Na (exchangeable), Na (ammonium lactate), Na (acid digestion),
Ni (acid digestion), Na (XRF), Zn (BaCl2 ), Zn (DTPA), Zn (Mehlich I), Zn (Mehlich
III), Zn (acid digestion), Zn (XRF).
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

• Other elements and mineralogy


Al (CaCl2 ), Al (dithionite), Al (exchangeable), Al (Mehlich I), Al (Mehlich III), Al
(oxalate), Al (oxalate), Al (water/H2 SO4 ), Al (acid digestion), Al (XRF), Al saturation,
Ba (XRF), Be (acid digestion), Ce (acid digestion), Ce (XRF), Cs (acid digestion), Cs
(XRF), Ga (acid digestion), Ga (XRF), gibbsite (termogravimetry), gibbsite (H2 SO4 ),
goethite (H2 SO4 ), H (exchangeable), H (oxalate/oxalic acid), hematite (H2 SO4 ), Hf
(XRF), In (acid dig), kaolinite (thermogravimetry), kaolinite (H2 SO4 ), kaolinite/gibbsite,
La (acid digestion), La (XRF), Nb (XRF), Rb (XRF), Sb (XRF), Sc (XRF), Se (acid
digestion), Si (oxalate), Si (water/H2 SO4 ), Si (XRF), Sr (XRF), Th (acid digestion), Th
(XRF), Ti (water/H2 SO4 ), Ti (XRF), Tl (acid digestion), V (XRF), Y (XRF), Zr (XRF).
• Electrical conductivity
• pH and lime requirement
Acidity (exchangeable), pH (CaCl2 ), pH (Mehlich buffer), pH (single buffered method),
pH (KCl), pH (water), lime buffer capacity, lime requirement.
• Soil contaminants
Atrazine absorption, diuron (Kd), benzopyrene (total), 17β estradiol (Kd ), polycyclic
aromatic hydrocarbon (tot), petroleum hydrocarbon (total), As (acid digestion), As
(XRF), Cd (acid digestion), Cr (acid digestion), Cr (XRF), Hg (acid digestion), Hg
(direct AAS-AMA254), Hg (fluorescence), Ni (XRF), Pb (BaCl2 ), Pb (acid digestion),
Pb (XRF).

Biological properties
• Microbial biomass
Biomass C, biomass C13, biomass C/C (organic), biomass C/biomass N, biomass
C/biomass P, biomass C/C (organic), biomass N, biomass N15, biomass P, substrate
induced respiration.
• Microbial respiration
C (dissolved-organic) production, C mineralization (basal), C mineralization (rate), C
mineralization rate (basal), C mineralization 50d, C mineralization 53d, C mineralization
220d, C mineralization 75 220d, C mineralization 264d, C mineralization 300d, C
mineralization 500d, C mineralization (basal)/N mineralization (rate), Co (potentially
mineralizable), metabolic quotient, N (active), N (mineralizable), N (potentially
mineralizable), N mineral (release), N mineralization 10d, N mineralization 50d, N
mineralization 53d, N mineralization 220d, N mineralization 75 220d, N mineralization
264d, N mineralization (rate), nitrification (potential).
(Continued on next page)
150 J. M. Soriano-Disla et al.

Table 2
List of soil variables predicted by reflectance spectroscopy together with chemometric
techniques as reported in the literature (Continued)

• Microbial groups
Actinomycetes, bacteria, bacteria (G), bacteria (G+), bacteria (G+)/bacteria (G),
bacteria abundance, bacterial activity, ergosterol, ergosterol/biomass C, fatty acid
18:2w6, fungi, phospholipid fatty acids (total), fungi/bacteria, protozoa,
vesicular-arbuscular mycorrhizal fungi.
• Enzymatic activities
Acid phosphatase, arylsulfatase, cellulase, dehydrogenase, denitrifying (activity),
denitrification (potential), FDA hydrolase, β-glucosidase, β-glucosaminidase,
nitrification (potential), peptidase, phosphatase, urease, xylosidase.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

results become more reliable with the number of published articles and determinations
available. Predictions based on R2 values were classified according to Malley et al. (5)
as follows: excellent (R2 > 0.95), successful (0.95 ≤ R2 ≤ 0.90), moderately successful
(0.90 ≤ R2 < 0.80), and moderately useful (0.80 ≤ R2 < 0.70). This classification offers
a simple guidance to interpret and compare the models obtained. However, the useful-
ness of a particular prediction depends on the nature of the analyte and potential use of
the predicted value (e.g., routine soil testing, mapping, and inputs into soil-related pedo-
transfer functions), as well as on the error and robustness of the prediction and relevance
of samples used for model development with regard to the unknown samples. In some
cases, models with R2 = 0.5–0.7 can be useful to assist with an understanding of the broad
principles relating the analytical data and spectral features (and thus soil chemistry), but
predictions with R2 < 0.5 are considered unreliable.
A majority of studies on the prediction of soil properties used the MIR and NIR spectral
regions, with a similar number of properties predicted using specifically the MIR, NIR, and
Vis-NIR regions (Table 3). Relatively few studies reported using the combined information
from UV-Vis-NIR, and even fewer with only the UV or Vis or the combined NIR-MIR
and Vis-NIR-MIR regions. Thus, almost all of the soil property predictions presented in
Table 3 were for the MIR, NIR, and Vis-NIR spectral regions. Consequently, the majority
of results and discussion of this review are centered on predictions using these spectral
regions, although some discussion is given for predictions using the Vis spectral region.
Predictions with the Vis region can be found in the studies of Viscarra Rossel et al. (10),
Bogrekci and Lee (78), and Nocita et al. (79), who all reported less successful predictions
than for other spectral ranges for a number of soil properties (e.g., the different forms of
P, TOC, pH, lime requirement, electrical conductivity [EC], and particle size). Moderately
successful calibrations, mainly for C fractions based on combined NIR-MIR with PLSR
modeling, have been reported by Ludwig et al. (80) and D’Acqui et al. (81), resulting
in median R2 = 0.85, and Shao and He (82) reported similar or worse results for major
nutrients (N, P, K) with either the NIR or MIR region. Predictions using the extended Vis-
NIR-MIR range did not result in improved accuracies compared to a single region tested
in the Vis, NIR or MIR (10).
The most common multivariate method used for deriving regression models from
spectral information and reference analytical data was PLSR. The number of samples for
model development ranged from about 20 to more than 18,000, with both cross-validation
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Table 3
Median R2 values (number of predictions in brackets) of validation results for prediction of soil properties using UV (100–400 nm,
25,000–40,000 cm−1), Vis (400–700 nm, 14,286–25,000 cm−1), MIR (2,500–25,000 nm, 400–4,000 cm−1), and NIR (700–2,500 nm,
4,000–14,286 cm−1) reflectance spectroscopy
UV-Vis- Vis-NIR-
MIR NIR Vis-NIR NIR-MIR NIR MIR UV Vis References
1. Physical properties
Aggregation Aggregation 0.79 (4) 0.59 (8) (71, 96, 97)
Soil water Moisture 0.83 (7) 0.86 (10) 0.86 (6) 0.89 (5) (3, 55, 71, 84, 86, 87, 94, 96, 98,
Field capacity 0.87 (2) 0.86 (1) 100, 101, 104, 107, 112, 118, 133,
194, 196–199)
Soil particle size Clay 0.80 (14) 0.72 (12) 0.78 (15) 0.55 (1) 0.72 (3) 0.67 (1) 0.43 (1) (3, 7, 10, 22, 31, 39, 41, 55, 56,
Sand 0.83 (11) 0.58 (8) 0.78 (8) 0.41 (2) 0.75 (1) 0.47 (1) 71, 81, 85, 91, 96, 97, 101–104,
Silt 0.63 (8) 0.46 (7) 0.67 (9) 0.10 (2) 0.52 (1) 0.31 (1) 110, 117, 118, 127, 129, 149, 156,
157, 162, 200–206)
2. Chemical properties
Carbon and organic C organic 0.88 (21) 0.88 (13) 0.78 (15) 0.85 (7) (3, 7, 10, 17, 20, 22, 31, 33–35,
matter (fractions) 37, 39, 41, 42, 48, 55, 58, 71, 74,
C (inorganic) 0.97 (3) 0.87 (3) 0.83 (9) 0.96 (1) 79–81, 83, 86–89, 91–94, 96, 97,
C (organic) 0.93 (20) 0.85 (23) 0.83 (33) 0.94 (1) 0.76 (4) 0.72 (1) 0.77 (3) 100–104, 107, 108, 110–112,
Organic matter 0.94 (3) 0.79 (9) 0.86 (14) 116–118, 127, 128, 131–134, 136,
C (total) 0.93 (16) 0.89 (24) 0.89 (27) 0.86 (1) 0.73 (3) 140, 142, 145, 149, 154–157, 160,
C/N 0.84 (5) 0.82 (7) 0.89 (8) 0.96 (1) 162, 163, 174, 175, 178, 179, 181,
182, 185, 187, 192, 194–196,
200–240)
Cation exchange Cation 0.85 (13) 0.81 (9) 0.84 (8) 0.77 (1) 0.71 (4) 0.09 (1) 0.16 (1) (3, 7, 10, 22, 39, 41, 55, 71, 81,
capacity exchange 94, 96, 100, 102–104, 110, 111,
capacity 116–118, 127, 156, 203, 204, 206,
210, 241)
(Continued on next page)

151
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

152
Table 3
Median R2 values (number of predictions in brackets) of validation results for prediction of soil properties using UV (100–400 nm,
25,000–40,000 cm−1), Vis (400–700 nm, 14,286–25,000 cm−1), MIR (2,500–25,000 nm, 400–4,000 cm−1), and NIR (700–2,500 nm,
4,000–14,286 cm−1) reflectance spectroscopy (Continued)
UV-Vis- Vis-NIR-
MIR NIR Vis-NIR NIR-MIR NIR MIR UV Vis References
Macronutrients Ca (exchange- 0.82 (8) 0.75 (11) 0.80 (7) 0.72 (4) 0.37 (1) 0.31 (1) (3, 10, 17, 20, 22, 33, 39, 42, 48,
able) 55, 59, 60, 71, 74, 78, 80, 82, 83,
Ca (total) 0.79 (2) 0.78 (6) 87, 90, 92–94, 96, 97, 100–103,
Ca/Mg 0.63 (1) 0.79 (2) 0.65 (1) 107, 110–112, 116, 118, 121, 122,
K (exchange- 0.37 (12) 0.61 (11) 0.71 (14) 0.72 (1) 0.32 (4) 0.46 (1) 0.29 (1) 127–129, 131–137, 140–142, 144,
able) 145, 154–156, 160, 174, 175, 178,
K (total) 0.70 (5) 0.50 (1) 0.75 (4) 179, 181, 182, 185, 187, 192, 194,
Mg (exchange- 0.74 (7) 0.73 (9) 0.83 (14) 0.57 (4) 195, 201–206, 208–210, 212, 214,
able) 216–219, 223, 225, 229, 231, 232,
Mg (total) 0.84 (4) 0.92 (1) 0.61 (6) 235–238, 240)
N (fractions) 0.72 (4) 0.80 (8) 0.85 (5) 0.77 (1)
N (mineral 0.85 (5) 0.45 (4) 0.57 (7) 0.02 (1) 0 (1)
NO3 /NH4 )
N (organic) 0.86 (1) 0.37 (2)
N (total) 0.90 (14) 0.86 (24) 0.86 (31) 0.83 (1) 0.69 (2)
P (extractable) 0.35 (11) 0.48 (14) 0.59 (27) 0.67 (6) 0.07 (1) 0.12 (2) 0.61 (3)
P (sorption) 0.83 (4) 0.58 (1) 0.65 (2)
P (total) 0.58 (5) 0.84 (2) 0.75 (4) 0.20 (1) 0.83 (1)
S (total) 0.43 (2) 0.84 (3)
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Micronutrients Co (total) 0.78 (2) 0.79 (2) 0.80 (1) (3, 20, 41, 42, 71, 74, 87, 96, 100,
Cu (extractable) 0.17 (2) 0.25 (1) 0.60 (2) 101, 107, 110, 111, 118, 122,
Cu (total) 0.82 (5) 0.68 (2) 0.67 (7) 127–129, 140, 144–150, 154–156,
Fe (extractable) 0.67 (6) 0.71 (2) 0.76 (4) 0.52 (1) 164, 175, 210, 221, 223, 234, 242)
Fe (total) 0.93 (5) 0.80 (3) 0.74 (8)
Mn (extractable) 0.57 (7) 0.55 (2)
Mn (total) 0.67 (2) 0.64 (2) 0.71 (3)
Na (exchange- 0.32 (6) 0.64 (7) 0.61 (2) 0.35 (4)
able)
Na (total) 0.71 (2) 0.50 (1) 0.81 (1)
Ni (total) 0.94 (3) 0.88 (3) 0.81 (3)
Zn (extractable) 0.20 (2) 0.46 (2) 0.670 (2)
Zn (total) 0.89 (4) 0.73 (2) 0.68 (8)
(Continued on next page)

153
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

154
Table 3
Median R2 values (number of predictions in brackets) of validation results for prediction of soil properties using UV (100–400 nm,
25,000–40,000 cm−1), Vis (400–700 nm, 14,286–25,000 cm−1), MIR (2,500–25,000 nm, 400–4,000 cm−1), and NIR (700–2,500 nm,
4,000–14,286 cm−1) reflectance spectroscopy (Continued)
UV-Vis- Vis-NIR-
MIR NIR Vis-NIR NIR-MIR NIR MIR UV Vis References
Other elements Al (exchange- 0.64 (11) 0.61 (3) 0.78 (3) 0.37 (1) 0.01 (1) (3, 10, 20, 71, 74, 101, 122, 127,
able) 140, 144, 145, 154–156, 203, 210,
Al (total) 0.87 (4) 0.46 (2) 0.78 (3) 221)
Si (extractable) 0.82 (2)
Si (total) 0.88 (2) 0.83 (1) 0.68 (1)
Ba, Be, Ce, Cs, 0.57 (51) 0.24 (9) 0.93 (1)
Ga, Hf, In, La,
Nb, Rb, Sb,
Se, Sr, Th, Tl,
V, Y, Zr (total)
Mineralogy Calcium 0.95 (3) 0.69 (3) 0.75 (4) (74, 86, 101, 102, 104, 155–157,
carbonate 206, 221)
Clay minerals 0.79 (2)
Fe/Al oxides 0.69 (4) 0.61 (3)
Electrical conductivity Electrical 0.26 (7) 0.57 (5) 0.60 (1) 0.15 (2) 0.29 (1) 0.05 (1) (3, 10, 71, 94, 100, 110, 118, 127,
conductivity 210, 221)
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

pH and lime requirement pH 0.75 (18) 0.57 (21) 0.79 (15) 0.67 (1) 0.72 (6) 0.73 (3) 0.36 (3) (3, 7, 10, 17, 20, 22, 31, 33, 39,
Lime 0.81 (2) 0.44 (4) 0.74 (1) 0.25 (1) 55, 58, 71, 81, 87, 96, 100, 102,
requirement 110–112, 117, 118, 127, 128, 142,
154, 156, 160, 162, 185, 201,
203–205, 210, 221, 225, 234, 238)
Pollutants Pesticides/ 0.72 (3) 0.40 (1) (3, 72, 74, 122, 144, 146–150,
steroids Kd 161–164, 166–168, 175, 221, 234,
Benzopyrene 0.90 (2) 242)
and PAH
Petroleum hy- 0.93 (1) 0.78 (2) 0.70 (3)
drocarbons
As, Cd, Hg, Pb 0.83 (13) 0.72 (6) 0.70 (26)
(total)
3. Biological properties
Microbial biomass Biomass C 0.84 (3) 0.82 (7) 0.82 (8) 0.61 (1) (3, 7, 17, 33, 80, 96, 100, 133,
Biomass N 0.82 (2) 0.70 (4) 0.93 (5) 136, 145, 160, 173–176, 179, 181,
Biomass P 0.62 (1) 0.81 (1) 185, 237, 240)
Microbial respiration C mineraliza- 0.82 (5) 0.76 (15) 0.77 (1) (7, 17, 33, 55, 80, 96, 100, 116,
tion 136, 145, 160, 174–176, 181, 185,
N mineraliza- 0.64 (3) 0.79 (7) 0.74 (10) 0.67 (1) 0.82 (1) 187–189, 212, 216, 237)
tion
Microbial groups Microbial 0.88 (1) 0.75 (14) 0.82 (2) (100, 133, 160, 173)
groups
Enzyme activities Enzymatic 0.70 (9) 0.80 (15) 0.78 (8) (20, 33, 100, 136, 172–174, 176,
activities 178, 182, 192, 243)

155
156 J. M. Soriano-Disla et al.

and test set validation equally used. The two most studied geographic areas are the United
States and Australia.
Only about 10% of the references cited in this review related to the use of portable
devices and in-field soil analysis (mostly in the Vis-NIR and NIR). Only one study (48)
reported results for a semiportable MIR device (the SOC-400 by Surface Optics Corp.).
This is largely due to the historically poor portability of MIR instruments in the commercial
market and the potential adverse effects of highly variable amounts of soil water on pre-
dictions from MIR spectra (15). However, a promising new range of true handheld diffuse
reflectance and internal reflectance FT-MIR spectrometers is currently being marketed by
Agilent (ExoScan and FlexScan series) that are suitable for in-field soil analysis.
The soil properties that have most extensively mainly been analyzed using portable
Vis-NIR and NIR devices under field conditions are extractable P, clay content, lime
requirement, concentrations of macronutrients (Ca, K and P), soil water content, TOC,
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

SOM, pH, and total C and N (42, 48, 56, 79, 83–94). When one measures in situ in
the field, one must account for water and other environmental effects, which are greatly
variable. Therefore, the comparisons on performance need to account for this. The median
R2 values for variables predicted for in situ soil samples were (for those variables with at
least three prediction models reported in the literature) 0.72 (extractable P), 0.78 (clay),
0.80 (TOC), 0.82 (soil water content), 0.88 (total C), and 0.84 (total N). The median
R2 values for the same variables, but using spectral information obtained with laboratory
instrumentation (generally using dried and sieved soils) were 0.52 (extractable P), 0.76
(clay), 0.85 (TOC), 0.86 (soil water content), 0.90 (total C), and 0.86 (total N). Comparisons
are not robust (e.g., for extractable P) due to the low number of studies examining in-field
calibrations.

Physical Properties
Physical properties are often seen as crucial for monitoring of the soil environment and
proper implementation of land management processes, but their determination is often
lengthy and expensive, leading to underutilization or lack of demand for such data. The
prediction of physical properties relies on their correlation with certain soil components
such as quartz (sand) and various types of clays, organic matter, carbonates, and oxides that
are infrared active.
The median R2 statistics of studies reporting predictions of soil physical properties are
presented in Table 3. In general, there were similar median R2 values for the Vis-NIR and
MIR, (e.g., R2 ∼0.80 for clay and sand). Attempts to use other regions of the spectra—that
is, UV-Vis-NIR and Vis-NIR-MIR—resulted in similar (e.g., for soil water content) or less
accurate predictions compared to the MIR or Vis-NIR.
Laboratory measurements of physical properties can be adversely affected by the
presence of soil carbonate minerals, sample heterogeneity, gravel, a variety of cracks and
voids in the samples, and compaction due to vehicle traffic. There is also the use of varying
analytical methods for the same analyte (e.g., combustion versus Walkley and Black for
soil organic C), drying of the samples during time of collection to final analysis, and very
wet samples close to or exceeding the field capacity, leading to reflection from water films
on soil surfaces. Standardization and great care in laboratory methods are thus critical.

Aggregation. Soil physical properties, including soil structure parameters, are often de-
pendent on the distribution and constitution of water-stable soil aggregates with a range of
sizes from 1 to 10 mm in diameter (95). Soil structure can influence many important soil
Soil Properties 157

properties and processes such as compaction, erodibility, water infiltration and retention,
and aeration. Some parameters related to soil aggregation that have been predicted include
aggregation (%) at 0.2, 0.25, 0.50, 1 mm and macroparticle size, R2 from 0.46 to 0.60, using
NIR; nonerodible aggregation (71), with MIR showing an R2 = 0.53; aggregation stability
(97), with NIR and MIR giving R2 = 0.92 and 0.80, respectively, and mean aggregate
diameter (R2 for NIR and MIR for prediction of mean geometric diameter = 0.67, 0.79,
and for mean weight diameter = 0.66, 0.79, respectively). Globally, MIR was superior to
NIR for the prediction of soil aggregation, with moderately useful predictions (median R2
= 0.79, 0.59, respectively, for the MIR and NIR), but only three reports were available for
these properties (71, 96, 97).

Soil Water. Knowledge of the water content of soil is critical for the development of
irrigation systems for successful plant growth and land management (84, 98). The prediction
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

of soil water content by spectroscopic methods is dependent on major soil components,


such as heavy clays (e.g., smectites and illites), SOM, and sand content (quartz). These soil
components affect the amount of water that the soil can retain at various suctions through
the formation of voids and fissures within and between microaggregates (3).
Reflectance spectroscopy can be a feasible alternative for the prediction of gravimetric
soil water content, because a considerable number of authors (see Table 3) have reported
predictions with moderately successful results for all spectral ranges: MIR, NIR, UV-
Vis-NIR, and Vis-NIR (median R2 values from 0.83 to 0.89, and as high as 0.98 for the
UV-Vis-NIR, 0.97 for the NIR, and 0.91 for the MIR). There was a slight advantage using
the NIR region (median R2 = 0.89 for the UV-Vis-NIR, 0.86 for the NIR and Vis-NIR,
and 0.83 for the MIR), probably not necessarily significant when considering the natural
variability in soil properties across a landscape.
Soil water content at field capacity is largely dependent on large pore spaces between
sand-sized particles and is well correlated with infrared spectra. An acceptable prediction
of soil water content at field capacity was found in the study by Viscarra Rossel and Webster
(43) using Vis-NIR (RPD = 1.7, equivalent to an R2 of 0.65) (99). A very high accuracy for
the prediction of this parameter was shown using the NIR (R2 = 0.86, RPD = 2.7) (100)
and the MIR regions (in Janik et al. (71) and Bertrand et al. (101), with R2 of 0.92 and 0.81,
respectively). However, additional confirmatory work is needed, because so few studies
have been reported. Soil water content at wilting point (1,500 kPa) is strongly dependent
on small pore spaces in clay-sized material and has been predicted well using MIR in the
study of Janik et al. (71) (R2 = 0.89) but with less accuracy using Vis-NIR (RPD = 2.0 and
equivalent R2 = 0.75) (43). The same study by Janik et al. (71) predicted water repellence
poorly (R2 = 0.25).

Particle Size Analysis. The prediction of particle size composition of soils with infrared
spectroscopy has been of ongoing interest, starting as early as 1995 (102) and has continued
to the present (103), resulting in a large number of predictions available (Table 3). Particle
size distribution influences soil texture and thus key soil processes such as soil water
dynamics and air diffusion controlling plant nutrition, microbial activity, and nutrient and
pollutant mobility. In addition, the clay content has a large influence on soil structure by
promoting the stability of soil aggregates and soil swell–shrink properties (8).
The presence of non-soil mineral material, such as SOM and carbonate, can result
in incorrect particle size measurements, so that steps (e.g., hydrogen peroxide and glacial
acetic acid treatments) must be used in the laboratory to remove such constituents that
otherwise would be included in the particle size fractions.
158 J. M. Soriano-Disla et al.

For clay and sand, almost equivalent R2 values have been obtained by using the
MIR (median R2 of 0.80 and 0.83, respectively) and Vis-NIR (median R2 = 0.78, 0.78,
respectively). The relatively high accuracy, the large number of research articles available,
and the fact that the calibrations rely on soil components such as quartz (sand) and various
types of clays that are infrared active, make infrared (especially MIR) a useful technique
for the prediction of these properties. On the other hand, lower values of median R2 were
generally observed for silt, probably as a result of uncertainty about the exact nature of soil
components associated with this large range of particle size, with the best results given by
Vis-NIR predictions (maximum R2 = 0.91, median R2 = 0.67). The MIR was slightly less
accurate (maximum R2 = 0.84, median R2 = 0.63).
Related to particle size, specific surface area was predicted by Ben-Dor and Banin
(104) and Ben-Dor et al. (86) using NIR and Vis-NIR, observing R2 prediction values of
0.70 and 0.92, respectively. Prediction of bulk density, another soil physical property with
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

agronomic importance, was moderately useful in Viscarra Rossel and Webster (43), who
observed an RPD = 1.9 (equivalent to an R2 = 0.72) in a set of 1745 Australian soils. A
slightly lower R2 (0.62) was found in an MIR study by Merry and Janik (105) using 1,124
Australian soils.

Chemical Properties
Predictions for soil chemical properties are by far the most common soil attributes reported
in the literature and are summarized below. Median R2 values are presented for predictions
of chemical soil properties using the different spectral ranges (Table 3).

Soil Carbon and Organic Matter. The quantity and nature of soil C pools often affect other
soil properties or characteristics such as pH, fertility, plant available water capacity, and
clay flocculation (106). Intensive and reliable mapping is thus required to monitor changes
in soil C pools (13). The monitoring of TOC contents in soil is gathering momentum as a
consequence of global warming and the impending implementation of C pricing policy. In
addition, TOC has been regarded as a useful indicator of sustainability and the status of soil
and plant nutrition parameters (3). The determination of inorganic C can be useful inputs
to soil classification and pedogenic process studies.
Infrared analysis is well suited for SOM analysis because of its sensitivity to the C H,
C O, and C N functional groups that dominate in organic matter. Soil organic matter
properties are most frequently estimated by Vis-NIR (8, 10, 107, 108), although the MIR
region was first reported for soil C predictions by PLSR (109), followed by further reports
including predictions of C pool chemistry (3, 33, 35, 108). Apart from soil organic matter, C
also occurs in the form of inorganic carbonates. Carbonates, widely occurring in many soils
throughout the world, give characteristics bands in the MIR (16) and NIR (41) (Table 1).
According to median R2 values in Table 3, MIR predictions for C (inorganic), TOC,
and SOM were reported to perform better than by NIR and Vis-NIR. However, median
values of R2 were similar for organic C pool fractions and total C using the MIR and NIR
frequencies, but in some cases predictions for C/N were improved using the Vis-NIR. The
prediction accuracies using the MIR were considered as moderately successful (organic C
fractions and C/N), successful (TOC, SOM, and total C), and excellent (for inorganic C).
Attempts to perform calibrations with extended NIR-MIR ranges for organic C fractions
(80) and UV-Vis-NIR for TOC (110) and total C (111, 112) did not improve the results
obtained with single regions (see Table 3). Thus, MIR appears to be the most successful
Soil Properties 159

region for the accurate prediction of the reported forms of C. Comparing NIR and Vis-NIR,
similar accuracies were observed except for SOM.

Cation Exchange Capacity. Cation exchange capacity (CEC) is an important soil parameter
implicated in soil fertility. It affects the release and retention of electrically charged nutrients
and clay interlayer cations and reflects the soil buffering capacity (8). Cation exchange
capacity is affected by SOM, particle size, surface area, and clay type (particular 2:1 clays
such as smectites). Specifically, interstitial water molecules surround the interlayer cations
Ca+2, Mg+2, K+, and Na+, and broad infrared peaks characteristic of this water occur in
the MIR region near 2450 to 3350 cm−1 (4,082 to 2,985 nm) and in the NIR near 2,000 nm
(5,000 cm−1) (3). There is an equilibrium between atmospheric relative humidity and
interstitial water content, with strongly bonded water molecules surrounding the cations
and hydrogen-bonded water molecules filling the spaces between the cation-bonded water
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

molecules. This results in either a single or double layer of water molecules dependent on
the hydration potential of cations (113, 114). Furthermore, there is a relationship between
the bonding energy of the cation-bonded water molecules and the nature of the cation (115).
There is thus a strong relationship between the spectral signatures of the water molecules
in clays and CEC.
The actual success of MIR prediction for CEC, reported in the literature, demonstrates
inconsistent results, with R2 ranging from 0.34 to 0.98 (median = 0.85), which may be
related to the wide range of chemical methods used in the laboratory to determine CEC. Low
R2 values resulted from an insufficient number of samples representing specific locations
(10, 116, 117). Many of the other cases of predictions could be classified as successful or
even excellent, with R2 > 0.92. Successful predictions were not unexpected in the MIR,
because this region is very sensitive to many of the soil components that correlate with
CEC. Values for the coefficient of determination were similar for NIR and Vis-NIR but
there were a lower number of predictions reported. There are three particular studies (110,
111, 118) that extended the range to the UV-Vis-NIR, but they only achieved a median R2
value of 0.71, considerably lower than that obtained with MIR, NIR, or Vis-NIR. Thus,
MIR, NIR, and Vis-NIR are useful for the accurate prediction of CEC, presenting a similar
accuracy with the highest number of reports found for the MIR region.

Elemental Analysis. It is well known that soil macro- and micronutrients play the most
important role in soil fertility (14). Predictions are feasible by means of the presence of these
elements as constituents in molecular groups (with characteristic vibration frequencies) that
absorb in the infrared region (e.g., the CO3 group in soil carbonates) or indirectly if the
element concentrations are correlated with other infrared-active compounds in the soil (e.g.,
Al, Fe, or manganese [Mn] oxides and oxyhydroxides).
Of all of the elements, only those present in soil components such as clays, Fe oxides,
hydroxides, and sulfates containing Al, calcium (Ca), Fe, potassium (K), magnesium (Mg),
sodium (Na), phosphorus (P), sulfur (S), silicon (Si), and titanium (Ti) can have Vis-
infrared–active bands. Plant nutrients are not expected to have direct absorption features in
the Vis-NIR region (8). For example, Ca2+, Mg2+, Al3+, Fe2+, and K+ are major cations in
carbonate minerals (e.g., calcite, dolomite) and in the interstitial spaces of 2:1 layer silicate
clays as cations (hydrated in heavy clays such as smectites or bentonites), thus also giving
strong signals in the MIR and Vis-NIR region (Table 1). Such cations are implicated in
cation exchange phenomena or clay structure bonding (3), contributing to the performance
of predictions for exchange properties.
160 J. M. Soriano-Disla et al.

There is great interest in estimating the potentially available fraction of nutrients in


soils through the use of different chemical reagents (see Table 2). This is the fraction most
likely to be taken up by plants and can assist in the assessment of the environmental fate and
transport of nutrients. Predictions of the available fraction of nutrients depend on the quality
of the analytical data, frequently adjusted, as is the case of available P tests as discussed by
Shepherd and Walsh (119), for local variation of soil properties, rendering the derivation of
calibrations for these extracted fractions on different soils challenging. This challenge has
contributed to the inability to obtain consistently good estimates across a range of soils,
relatively high standard errors, and significant effects of soil type (120). If these fractions
are related more to the concentration of the element in solution, rather than to the chemistry
of the soil matrix, then the infrared method will probably prove unsuccessful (3). In the case
of exchangeable cations, the interest is to have an indication of this fraction, because the ion-
exchange characteristics influence the physical and chemical properties of most soils (106).
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

The considerable heterogeneity of nutrient concentration on a small scale throughout


a cultivated field and the different nutritional demands of various crop varieties require a
relatively high sampling effort for the development of models for these fractions.

Soil Macronutrients (Ca, K, Mg, N, P, S). Considerable effort has been made in the deter-
mination of soil fertility macronutrient parameters by infrared spectroscopy particularly for
exchangeable Ca, K, Mg, extractable P, and total nitrogen (N). Determining soil macronu-
trient levels (especially for plant available N, P, and K) is important for increased crop and
plant production, enables optimization of fertilizers inputs, and helps to establish criteria
for application rate scheduling (14). Easy access and acquisition of such data could help
to lower agricultural production costs and reduce the potential for adverse environmental
impacts due to overapplication of fertilizer nutrients (120), leaching into groundwater, or
depletion of valuable nutrient stocks.
In general, the performance of regression models using the MIR, NIR, and Vis-NIR
spectroscopy has been shown to be similar for predictions of concentrations of soil macronu-
trients (median R2 for soil macronutrient regressions using the MIR, NIR, and Vis-NIR =
0.74, 0.77, and 0.72, respectively). The small differences in accuracy can be related to the
use of different soil tests used and different concentration ranges assessed (e.g., extracted
by different reagents, exchangeable and total). Considerably more data are available for the
prediction of exchangeable/extractable nutrients (e.g., Ca, K, Mg, and P, which are the most
important fractions in agronomic terms) than for total elemental concentrations. However,
it could not be concluded for which fraction the highest accuracies were obtained because
this depends on the particular nutrient element and the spectral region being considered, as
discussed above.
Predictions using the combined UV-Vis-NIR range were reported for exchangeable
Ca, Mg, K, extractable P, and for total N (90, 110–112, 118, 121). In general, the median
R2 values for predictions using the combined UV-Vis-NIR were no higher than those for
the NIR, MIR, or Vis-NIR ranges (Table 3). Pirie et al. (110) found that the MIR region
was better than UV-Vis-NIR for the prediction of exchangeable Ca and Mg (R2 = 0.64
and 0.69, RPD = 1.6 and 1.7 for the MIR compared to R2 = 0.28 and 0.28, RPD = 0.2
and 1.2 for the combined UV-Vis-NIR, respectively, for the two nutrients) and similar for
exchangeable K (R2 = 0.27, RPD = 1.1 for the MIR compared to R2 = 0.26, RPD = 1.2
for the combined UV-Vis-NIR).
Calcium. The primary source of exchangeable Ca is the presence of Ca2+ cations in
the interstitial layer of heavy clays such as smectites and the edges of other clay minerals
Soil Properties 161

(3, 29). This Ca can be displaced from the clay interlayers by other cations such as Na+ and
Mg2+. Some Ca may also be available as a result of the presence of calcite and dolomite
minerals in the soil (39). These sources of Ca are characterized by strongly infrared-active
materials and thus generally result in good prediction models.
The three main regions (Vis-NIR, NIR, and MIR) used for the prediction of exchange-
able Ca resulted in similar median R2 values (and similar number of determinations reported
in the literature), with the highest median value for the MIR (R2 = 0.82) and with individual
R2 values > 0.90 common. In this region, calcite (CaCO3 ) and heavy clays in the soils are
expected to contribute strongly to the exchangeable/extractable Ca models (109). For total
Ca, the majority of the determinations were available for the Vis-NIR region, resulting in
moderately useful predictions (R2 = 0.78). Two studies (3, 122) reported predictions of
total Ca using MIR resulting in a similar value for a median R2 of 0.79.
In summary, Vis and infrared spectroscopy (MIR, NIR, and Vis-NIR) can be used as a
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

surrogate for the prediction of soil Ca (exchangeable and total).


Potassium. Potassium is a primary macronutrient for plants, sometimes required in
quantities equivalent to and sometimes greater than that of N (123). The most abundant
sources of K in mineral soils are K-feldspars and micas, and the weathering of the latter may
produce secondary minerals (illite and many mixed layer clays) that in turn also represent
potential sources of K for plant nutrition (124). Compared to Ca2+ cations, K+ is usually
not hydrated and is held more strongly in the interstitial clay layer, thus assisting in binding
the clay aluminosilicate sheets together.
Poor results were observed for the prediction of exchangeable K using MIR (median R2
= 0.37). Better predictions, but still with low values of R2, were found for the NIR (median
R2 = 0.61). The best results were observed for the Vis-NIR (median R2 = 0.71). The better
results including the NIR region may be due to the absorbance by illites of NIR frequencies
(Table 1), whereas the spectra of smectites and illites are very similar in the MIR, apart
from the intensity of water absorptions near 3,300–3,450 cm−1 (2,899–3,030 nm). However,
moderately useful predictions were also reported for exchangeable K using the combined
NIR-MIR region (R2 = 0.72), but this result was from only a single study (82) showing a
similar performance for NIR (R2 = 0.69) and better for the MIR (R2 = 0.79). Site-specific
models for a single paddock may show better performance than global models for the
prediction of exchangeable K (125). However, such predictions cannot be transferred to
other sites. For the prediction of concentrations of total K, the best results were found for
MIR and Vis-NIR (median R2 value of 0.70 and 0.75, respectively).
It thus appears that infrared spectroscopy does not have sufficient information to
provide an accurate prediction of K in soils, especially for exchangeable K. Although total
K depends on soil mineralogy, the exchangeable fraction of K is influenced by factors
(e.g., the concentration of K+ in soil solution) that are not infrared active. Influenced by
applications of K as fertilizers and the fact that, in general, K is leached more readily than
Ca/Mg, a considerable large amount of exchangeable K can be in the form of soluble ions
(126). Potential issues related to the variability of nutrient availability (which requires a
higher sampling effort) and the use of different laboratory methods that do not necessarily
reflect the same K pool can also influence the poor results obtained. In addition, the
concentration of exchangeable K is relatively low compared to Ca and Mg in soils and the
effect of K+ on the soil spectra may be largely hidden or masked by those for Ca2+ and
Mg2+ (127).
Magnesium. Magnesium occurs as a major soil element along with Ca and K. The main
spectral signatures are expected to be due to Mg-Ca-carbonates (dolomite), as a hydrated
162 J. M. Soriano-Disla et al.

cation in heavy clays, and substituted for Al in the octahedral structures in micas (29, 126).
Infrared prediction results were similar in accuracy to that of Ca and far more accurate
than for K. High accuracy was found for the prediction of exchangeable Mg by using either
MIR (71) (R2 = 0.92), NIR (128) (R2 = 0.85, RPD = 2.6) or Vis-NIR (129) (R2 = 0.90),
with median R2 values for Vis-NIR (0.83) slightly higher than for MIR and NIR (0.74 and
0.73, respectively), as shown in Table 3. In the comparative study of Van Groeningen et al.
(116), Vis-NIR performed better than MIR for the prediction of exchangeable Mg (R2 =
0.82, RPD = 2.3 and R2 = 0.61, RPD = 1.5 for the Vis-NIR and MIR, respectively).
Magnesium concentrations assessed by acid digestion and X-ray fluorescence (XRF)
were predicted with similar accuracy using MIR (median R2 = 0.84), whereas poor results
were found for Mg concentrations extracted with acid digestion using Vis-NIR (median R2
= 0.61). Reeves and Smith (74) reported an R2 = 0.92 and RPD = 3.4 for the prediction
of the concentration of Mg extracted with aqua regia using NIR. This compares to an R2 =
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

0.96 and RPD = 5.1 obtained in the same study using MIR.
Spectroscopy can therefore be useful for the prediction of exchangeable and total Mg,
with Vis-NIR and MIR, respectively, showing the best results. According to the successful
results found for the prediction of total Mg in the study by Reeves and Smith (74) using
NIR, this region can be also potentially useful for the prediction of the total concentration
of this element.

Nitrogen. Nitrogen is one of the most important nutrients in agricultural cropping,


pasture, and horticultural systems together with P and K (126). A total of four groups of
properties related to N are found in Table 3: N fractions (N in particulate organic matter
and particle size fractions), mineral N, organic N, and total N (usually determined by
dry combustion and Kjeldahl method). In addition, N supply was predicted with poor to
marginal accuracy in the studies of Fox et al. (130) and St. Luce et al. (131) using Vis-NIR
and NIR, resulting in R2 values of 0.49 and 0.62, respectively. Soil N supply is a measure
of potentially available N from the soil for crop uptake during a growing season (131).
Similar high median R2 values were obtained for the prediction of total N using MIR
(median R2 = 0.90), NIR, and Vis-NIR (median R2 = 0.86), but the Vis-NIR was more
accurate for the N fractions (median R2 = 0.85). Despite the concentration of N in soil
generally being below 1%, N is a major element in SOM and is highly correlated with soil
C concentrations, thus providing the infrared method with a capability to predict total N
(11). In addition, amide groups in SOM constitute a major component of total N and are
strongly infrared active (Table 1), explaining the good results for total and organic N (3,
11, 132, 133).
As stated by Janik et al. (3), temporal changes in soil mineral N (nitrate and ammonium)
are important issues when developing infrared calibration for these variables. Predictions
for N mineral, using MIR-ATR spectroscopy (59), rather than the usual DRIFT technique,
could be regarded as successful (R2 = 0.86–0.96) (59). Moderately successful predictions
(R2 = 0.85) for N (KCl extractable) were also obtained by Du et al. (134) using PAS-
MIR spectra in a test set limited to 42 calibration samples and 12 validation samples.
Results for the prediction of mineral N using DRIFT and MIR are generally poor (10,
127). Soil nitrate concentration should be directly measured due to the relatively strong and
sharp -NO3 stretching vibration peak near 1350 cm−1 (7,407 nm) (59). According to Jahn
et al. (59), there is some interference from carbonate, and problems due to the relatively
low concentration of nitrate ion in most soils generally make these predictions unreliable.
Predictions for the Vis-NIR and NIR were carried out by the diffuse reflectance method
resulting in median R2 values of 0.45 but with R2 values as high as 0.79–0.81 (135, 136).
Soil Properties 163

In summary, Vis and infrared spectroscopy can be used for the prediction of dif-
ferent N forms, with MIR as the region that provides the highest R2 values for total,
organic, and mineral N and the Vis-NIR for the N fractions. It is important to point out
that, compared to total N, a small number of determinations were available for N fractions,
mineral, and organic and good results for mineral N prediction using MIR were provided
by ATR and PAS spectroscopy.
Phosphorus. Phosphorus is required in relatively large amounts by plants for root
growth and seed development (123, 137). Though data on total P concentration may be
useful for mass balance calculations in agricultural systems or in environmental studies, in
general the analysis of P in soils focuses on the determination of the potentially available
fractions and on the ability of soil to bind P (P sorption). The availability of P in soils is
often dependent on the extent of sorption of P by soil solid phases (mineral and organic
matter), particularly in very acidic and very alkaline soils (138, 139). There are a number
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

of standard analytical methods used to determine the various forms of extractable P in soil
(e.g., Bray, Colwell, Mehlich, and Olsen).
With a few exceptions of soil sets representing special or unusual conditions, most
predictions of extractable P in soils resulted either in low R2 values (0.5–0.7) or were
considered to be completely unreliable (R2 < 0.50). Such poor results may be partly due to
the fact that these tests are related more to the concentration of P in soil solution between
the soil aggregates rather than to the chemistry of the soil matrix (3).
However, good predictions were observed for P sorption with MIR as the region that
resulted in the best calibrations (moderately successful predictions, median R2 = 0.83) (3,
71, 127). Studies using the Vis-NIR were less successful (median R2 = 0.65) (140, 141).
Success in predictions of P sorption by infrared spectroscopy may be attributed to the
response of the infrared spectra to the chemistry of soil components in the soil matrix that
are known to present a great affinity for phosphate anions (i.e., certain minerals such as Al-
and Fe-oxyhydroxides and carbonates and SOM/Al complexes) (3).
Phosphate minerals are characterized by strong absorption bands in the MIR region
(142). However, in soil spectra, bands due to P compounds can overlap with absorption
bands due to other soil components (143). In the Vis-NIR, the phosphate ion is not spectrally
active (137). Thus, total P models for both regions depend on direct and indirect correlations
with infrared-active soil compounds. Low accuracies were reported for the prediction of
total P using MIR (122, 127). The Vis-NIR and NIR regions were more successful (median
R2 = 0.75 and 0.84 for the Vis-NIR and NIR, respectively), with the highest R2 found in
the studies by Cohen et al. (20) and Bogrekci and Lee (78).
Overall, predictions for extractable P have been generally poor and well documented
by the large number of studies examining this property (see Table 3). Better results were
obtained for P sorption and total P using MIR and Vis-NIR, respectively, but with fewer
supporting data (Table 3).
Sulfur. Very few studies on the prediction of S are available, but it is known that
total S is related to S in SOM, gypsum (hydrated CaSO4 ), clay minerals, hydrous Al and
Fe oxides, and the presence of CaCO3 (101, 126, 144). Predictions of total S were only
available for the MIR and Vis-NIR. Better median results were found for the latter (R2 =
0.84) (101, 145). Similar R2 values were obtained by Kemper and Sommer (144) using
MIR, which contrasts with the poor results (R2 = 0) found by Minasny et al. (127) for the
prediction of the same element assessed with the same method (XRF). We suspect that such
a discrepancy was related more to analytical problems rather than to the inadequacy of MIR
spectroscopy to predict the concentration of S. Malley et al. (107) predicted extractable S
164 J. M. Soriano-Disla et al.

(extracted in ammonium acetate and acetic acid) with moderate success using Vis-NIR (R2
= 0.85, RPD = 2.6). High accuracy for the prediction of extractable S is not unexpected
considering the importance of the S organic pool in soils, the relatively large fraction of this
pool that is labile (126), and the sensitivity of infrared to the organic compounds (Table 1).
Although the utility of spectral methods to predict extractable and total S in soils needs to
be confirmed by further research, the potential is unquestionable.

Soil Micronutrients (B, Cl, Co, Cu, Fe, Mn, Mo, Na, Ni, Zn). Compared to macronutrients,
very few articles are available reporting predictions of micronutrient concentrations in
soils by spectroscopic methods. With the exception of Fe, the performance of the infrared
calibrations for these elements is almost entirely dependent on their correlations with
other infrared-active soil constituents (101, 122). Micronutrients are generally in such low
concentrations in soil that, even if they have inherently strong spectral peaks, these would
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

probably not be observed in the spectra of soils at these low concentrations.


Of all of these elements, total copper (Cu), Fe and zinc (Zn), extractable Fe and Mn,
and exchangeable Na were those with a largest number of reported determinations and
are discussed in this section. Results have generally included the MIR, NIR, and Vis-NIR
spectral regions, but Islam et al. (118), Pirie et al. (110) and Mouazen et al. (111) attempted
to use UV-Vis-NIR for exchangeable Na with poor results (median R2 = 0.34).
In general, better prediction results were found for the total elemental concentra-
tions than for extractable and exchangeable fractions (Cu, Fe, Mn, Na, and Zn). Infrared
spectroscopy is unlikely to provide accurate predictions for soil determinations that are
dependent more on soil solution chemistry than the soil matrix (3). Fortunately, the concen-
trations of some elements in the soil (e.g., cobalt [Co], Cu, Mn, nickel [Ni], Zn) have been
found to be related to carbonate (101) and Fe minerals (122, 144, 146, 147), with the latter
being the main parameter determining the accuracy of infrared models for these elements.
Other soil infrared-active covarying properties that have been found to be important for the
prediction of these elements are clay minerals and SOM (148–150).
The prediction of total Na concentrations by infrared spectroscopy is challenging
because it can form NaCl, an infrared-inactive ionic salt. A large proportion of NaCl can be
dissolved in water films adhering to soil particle surfaces (151). In addition, natural sources
(rainwater and wind-transported materials) and inputs of pesticides, fertilizers, biosolids,
industrial wastes, and saline or sodic waters to agricultural soils can result in increased
concentrations of Na (152) that are not related to bulk soil constituents. Despite these
difficulties in deriving regression models, moderately useful predictions (R2 = 0.75, RPD
= 2) have been obtained for the prediction of concentrations of Na (determined by XRF)
in the study by Soriano-Disla et al. (122), observing negative peaks in the derived model
due to kaolinite and carbonate.
The MIR region resulted in the best predictions for total Cu, Fe, Ni, and Zn and
extractable Mn, with median R2 values of 0.84, 0.93, 0.96, 0.90, and 0.57, respectively.
However, for extractable Fe, the Vis-NIR predictions gave the highest R2 values (0.76),
and NIR was shown to be the best spectral range for exchangeable Na (median R2 of
0.64). Mid-infrared predictions were poor for extractable boron (B; R2 = 0.30) (3) and Na
absorption ratio (R2 = 0.32, RPD = 1.1) (127) and moderately useful for Cl (chloride; R2
= 0.76, RPD = 2.0) (127). Though infrared is only sensitive to the fundamental vibrations
of Cl− bonds at low frequencies (usually less than about 500 cm−1), a possible explanation
of the ability to predict Cl is that the Cl− ion modifies the H-bonding environment of the
water molecules according to concentration (153), and therefore the water frequencies,
thus possibly allowing its prediction. Furthermore, according to Minasny et al. (127), EC,
Soil Properties 165

exchangeable Na percentage, and chloride are found to be highly correlated, particularly in


Vertosols and Dermosols; thus, the calibration is often specific to soil types and particular
environments.

Other Elements. In addition to the usual macro- and microelements, Al, Si, and Ti are
also important soil compositional elements. Aluminum, Si, and Ti occur as components at
relatively high concentrations in many soil minerals and thus are expected to be amenable
to infrared spectroscopic analysis. Aluminum is a major component of the structure of
most soil minerals such as clays (along with Si), oxides, and oxyhydroxides and thus has
well-defined peaks in the MIR contributing to the successful predictions found for total Al
(median R2 = 0.87). The prediction of exchangeable Al was less accurate with Vis-NIR
showing the best results (median R2 = 0.78). For exchangeable Al using Vis-NIR, only three
predictions are reported (20, 140, 154) in contrast to the relatively large number available
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

for MIR (11). Exchangeable Al is produced by soil acidification and can adversely affect
root growth of many crops.
Silicon is an important structural component of primary and secondary crystalline
silicates and Al and Fe silicates. Quartz, a primary source of Si, is a major component
of many soils and is expected to show a strong positive correlation with Si concentration.
Furthermore, the quartz -Si-O frequencies are among the strongest intensities in the MIR,
with peaks dominating the frequencies from about 600–2,000 cm−1 (16,667–5,000 nm). In
clays, the primary Si-O bonds also absorb very strongly, particularly in the broad region near
1,020–1,040 cm−1 (9,615–9,804 nm) (29). Predictions of Si in soils have therefore usually
been successful, with results obtained using MIR ranging from moderately successful to
excellent (R2 = 0.82 and 0.97 for oxalate-extractable and total Si, respectively) but, again,
few determinations have been reported (3, 101, 127, 155). In contrast to MIR, NIR does not
appear to have any observable quartz frequencies, only indicating the presence of quartz
by scatter effects (8).
Titanium is an element widely distributed in the Earth’s crust present as a component
in heavy mineral sands (e.g., ilmenite, anatase, brookite, and titanite) and geological parent
materials, with very little contributed from anthropogenic inputs (29). Titanium is often
correlated inversely with Al and Si. Total Ti has been predicted using MIR, with R2 values
of 0.78 (3) and 0.69 (RPD = 1.8) (122).
In summary, the prediction results for Al, Si, and Ti are promising (especially for the
total concentrations using MIR), but further confirmation is still required. The prediction of
other elements including barium (Ba), beryllium (Be), cerium (Ce), cesium (Cs), gallium
(Ga), hafnium (Hf), indium (In), lanthanum (La), niobium (Nb), rubidium (Rb), antimony
(Sb), selenium (Se), strontium (Sr), thorium (Th), thallium (Tl), vanadium (V), yttrium
(Y), and zirconium (Zr) have been considered together, because only a few determinations
were reported and these elements usually only occur in minor concentrations in soils. The
best results were observed for MIR, although not particularly successful, due to the low
frequencies of their molecular vibrations and relatively low accuracies (median R2 = 0.57).
For these elements, there was only one prediction using Vis-NIR, reporting a high R2 of
0.93 for Sb (RPD = 4.7) (144).

Mineralogy. Soil mineral type, proportion, and concentration determine soil properties
such as texture, structure, and CEC (8) are important for soil classification (156) and are
responsible for P sorption (3). Spectral signatures for soil minerals in the Vis-infrared
region—for example, quartz, clay, and Fe oxides—are numerous and are summarized
in Table 1. However, despite this, only a limited number of studies have attempted to
166 J. M. Soriano-Disla et al.

quantitatively predict the mineralogical composition of soils. This is partly due to the fact
that conventional analytical methods, such as X-ray diffraction, are primarily used for the
clay fraction of soils, and data are often unavailable for the whole soil. In many other
instances, the X-ray diffraction method is used to provide qualitative data (56). These
authors showed that Vis-NIR can be used to rapidly estimate the main mineral components
of soils. Mineral identification by Vis-NIR can include those soil mineralogical components
that would not normally appear in X-ray diffractograms due to poor crystallinity (e.g.,
amorphous Fe oxides).
Calcium carbonate content is readily determined by conventional laboratory methods,
and NIR, MIR, and Vis-NIR have all been used to predict this property (86, 101, 104).
Predictions for MIR (median R2 = 0.95) were classified as successful, followed by the
Vis-NIR and NIR (R2 = 0.75 and 0.69, respectively). The paper by Vendrame et al. (156)
reported quantitative predictions for the mineralogical content of Brazilian soils using
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

NIR. Coefficient of determination and RPD values as high as 0.86 and 2.5 (gibbsite),
0.56 and 1.5 (goethite), 0.6 and 1.6 (hematite), 0.83 and 2.2 (kaolinite), and 0.83 and 2.2
(kaolinite/gibbsite) were obtained from a diverse set of soils. The abundances of kaolinite,
illite, and smectite were measured in Australian soils using NIR spectroscopy in the study
by Viscarra Rossel (24). The amount of Fe oxides in soils extracted by different reagents
(dithionite combined with either citrate and/or bicarbonate) has also been predicted by
Vis-NIR in the studies by Ben-Dor and Banin (155), Ben-Dor et al. (86), and Summers
et al. (157), who obtained R2 values of 0.61, 0.90, and 0.61 (RPD = 0.6), respectively.
Viscarra Rossel et al. (158) mapped the relative abundance of hematite and goethite and
soil color across Australia from their Vis-NIR spectra.
Electrical Conductivity. Electrical conductivity (commonly determined in a 1:5 soil :
water extract) is a widely used parameter for describing soil salinity. It is well known that
the concentration of electrically charged water-soluble salts influences plant growth and soil
structure (106). Soil EC can be used as a surrogate measure of salinity, soil water content,
topsoil depth, and clay content and provides important information about the impact that
farm practices, such as irrigation and soil and crop management, have at both the field and
regional scales (159).
Electrical conductivity is generally poorly predicted by both MIR (median R2 = 0.26)
and NIR spectroscopy (median R2 = 0.57). A similar accuracy (R2 = 0.60) was found in one
report for Vis-NIR (94). These results are in agreement with Viscarra Rossel et al. (10), who
stated that, generally, properties related to electrolyte concentration, such as EC and SAR,
cannot be predicted accurately because they are based on the soil solution rather than the soil
matrix. Minasny et al. (127) stated that EC of the soil is influenced by a wide range of factors,
sometimes not necessarily related to the soil properties but often by landscape and climatic
factors. The performance of predictions for EC (and also soil pH as discussed below) is
often specific to soil types and particular environments and depends on the relationships of
these variables with infrared-active compounds (e.g., SOM and clay) (100).
pH and Lime Requirement. Soil pH and lime requirement are important parameters in
agriculture, because key soil mechanisms such as nutrient solubility, biological activity, and
organic matter mineralization are strongly influenced by soil pH. Generally, a pH close to 6.5
is regarded as ideal and liming is often undertaken to ameliorate soils having lower pH (8).
Soil pH and lime requirement can be predicted using infrared spectroscopy through
correlations with other soil properties (i.e., SOM and clay). Soil acidity develops from
mineralogical sources, such as proton-rich clays, Al oxyhydroxide minerals and sulfides,
oxidizable ammonium and organic N (as amides), and organic acids (carboxylic acids
Soil Properties 167

and phenols) (3). The mechanisms responsible for the generation of soil acidity can vary
significantly from one soil type to another (8, 33). Thus, sample diversity due to different
locations and depth may affect the performance of pH calibrations. This is something to
consider when developing models for soil pH, because such indirect calibrations may lead
to instability problems for global calibrations over a large variety of soil types or over large
geographical areas (11).
Also important when dealing with assessment of pH predictions are the often favorable
values of RMSE reported (8). It is interesting to note that the median R2 and RMSE for
all predictions found for pH (water) are 0.72 and 0.29 pH units, respectively, and 0.70 and
0.30 pH units, respectively, for pH (CaCl2 ). Despite the relatively low R2 units, an RMSE
of 0.31 pH units may be sufficient to provide useful Vis-infrared models. This error can be
reduced using field or farm-specific calibrations.
There are reported predictions for pH in water using MIR where results can be classed
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

as moderately successful with an R2 ∼ 0.85 (33, 117), but most R2 values were < 0.70,
although these could still be useful due to the favorable RMSE values. Predictions for pH in
CaCl2 were slightly better, where moderately useful predictions were obtained for Vis-NIR
(up to R2 = 0.88) (160) and > 0.92 for MIR (117). For relatively large sets, the best results
were found using Vis-NIR and MIR (median R2 = 0.79 and 0.75, respectively). Attempts to
use other spectral region combinations (UV-Vis-NIR or Vis-NIR-MIR) have not improved
prediction accuracy. Thus, soil pH can be successfully predicted due to indirect correlations
with other soil parameters, although such correlations can change from one set of soil to
another requiring the development of local regression models.
With regard to the amelioration of soil acidity, the amount of limestone required to raise
the pH of an acidic soil by 1 unit is known as the lime requirement (tonnes/ha/0.1 m depth).
Care must be taken not to overlime; otherwise, the excess lime can lead to trace element
immobilization. Lime requirement is dependent on soil pH, organic matter. and clay/sand
content and therefore should be predictable by infrared spectroscopy. The best results were
found for the MIR (with predictions considered as moderately successful) (3, 10). Janik
et al. (3) suggested that the ability to predict lime requirement depended on a number of
factors, notably clay type and content, Al- and Fe-oxyhydroxides, organic matter content,
and the negative influence of sand (as quartz) (33). Janik et al. (3) also stated that prediction
uncertainty could be attributed more to high variance in the laboratory analytical data rather
than to an inherent inability of the infrared technique to predict lime requirement.

Soil Contaminants. Soil contamination can result from agricultural chemicals and from ur-
ban and industrial activities. Of these contaminants, total petroleum hydrocarbons (TPHs),
total polyaromatic hydrocarbons (PAHs), benzopyrene, 17βestradiol, potentially toxic met-
als (arsenic [As], cadmium [Cd], lead [Pb], and mercury [Hg]) and organic pesticides such
as atrazine and diuron have been modeled by MIR, NIR, and Vis-NIR spectroscopy. The
presence of these contaminants in soils is strictly regulated because they can have adverse
health and environmental impact. In some cases, the contaminant concentrations are deter-
mined, and in other cases, such as for metals, pesticides, and steroid hormones, the retention
in soil (measured by partitioning Kd or sorption coefficients) is required. A considerable
number of prediction values are available for potentially toxic metals in comparison to
other contaminants.
Despite the fact that metals are not expected to be infrared active, generally good
results for total metal concentrations by aqua regia have been reported, with the best results
found for MIR (median R2 = 0.88). The concentration of these metals can be predicted due
168 J. M. Soriano-Disla et al.

to their correlation with other infrared-active compounds, that is, carbonates, clay minerals,
Fe oxides, and SOM (144, 146–150).
The fate of organic pollutants (such as pesticides and steroid hormones) in soil is
controlled by their sorption onto the soil matrix (161, 162) expressed by the Kd . The Kd
is dependent on SOM and mineral composition, both capable of being predicted from
infrared regression models. Similar moderately useful predictions for atrazine sorption
were obtained in the studies by Janik et al. (3) and Kookana et al. (161) using MIR (R2
= 0.69 and 0.72, respectively). Forouzangohar et al. (163) showed how MIR (R2 = 0.81,
RPD = 2.3) outperformed Vis-NIR (R2 = 0.40, RPD = 0.4) for the prediction of diuron Kd .
However, successful predictions of 17β-estradiol Kd were obtained by Singh et al. (162)
using NIR (R2 = 0.93, RPD = 3.8).
Prediction calibrations for concentrations of benzopyrene and PAH have been reported
for the Vis-NIR (164), resulting in successful predictions (R2 = 0.90). Janik et al. (165)
have also shown accurate prediction of PAH concentrations at an RMSE of <300 mg kg−1
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

for test soils using aromatic peaks in the MIR.


Petroleum hydrocarbons have direct absorption in the infrared spectra, but the precision
of prediction results depends very strongly on the performance of the spectrometer (noise
level) and reliability of the TPH analytical data (166). In the case of TPH predictions,
Vis-NIR calibrations were reported to give median R2 and RPD values of 0.70 and 2.0,
respectively (72, 167, 168). This performance was similar to that of NIR with R2 and
RPD values of 0.72 and 1.8, respectively (166), and work by Forrester et al. (169) showed
excellent predictions using the MIR with both benchtop and handheld instruments.
Good results for MIR have been recently confirmed by Forrester et al. (170) using
Australian soils polluted with TPH. This study demonstrated that the use of DRIFT spec-
troscopy with PLSR was able to provide viable and accurate models for the prediction of
TPH concentrations, with the MIR range outperforming the NIR range. The PLSR regres-
sion analysis, using the combined 2,950–2,980 and 2,650–2,777 cm−1 (3,356–3,390 and
3,601–3,774 nm) MIR frequency region for the 0–15,000 mg kg−1 set, resulted in R2 = 0.93
and RPD = 3.7 (successful) and RMSE = 564 mg kg−1. The PLSR regression for the same
concentration range, but using the NIR range (4,120–4,540 cm−1; 2,203–2,427 nm) resulted
in RPD of only 2.4 and R2 = 0.84 (moderately successful) with an RMSE = 853 mg kg−1.
In summary, Vis and infrared have shown promising potential for the prediction of soil
contaminants, especially for total concentrations and sorption characteristics of potentially
toxic heavy metals (As, Cd, Hg, Pb), pesticides, hormones, benzopyrene, and concentrations
of total PAH and TPH.

Biological Properties
The microbial biomass is responsible for 80–90% of biochemical transformations in the
soil (171) and is central to C and nutrient cycling and energy flow in soil systems. Interest
in developing infrared-based predictions of biological indicators lies largely in the need for
rapid and cost-effective methods for monitoring and modeling soil quality at the landscape
level, where sample numbers are often large. Though the microbial biomass encompasses
wide diversity, it typically represents 5% or less of TOC content (3, 172). Consequently,
the spectral contribution from the soil microbial biomass (SMB) per se is not expected to
result in observable patterns in soil spectra (20, 133, 173) and in any case would be masked
by soil mineral and other peaks at these concentrations (3), especially in the MIR region
(mostly strong overlap with quartz). Predictions are thus likely to be a result of strong
correlations between the biological properties and the quantity (133, 173, 174) and quality
Soil Properties 169

of SOM (175). However, and according to Reeves et al. (172), Cécillon et al. (176), and
Zornoza et al. (100), correlations with C and N may only partially explain the infrared
PLSR predictions of biological properties, with differences in macronutrient availability
(amount and quality) and specific signature wavelengths directly related to the microbial
biomass (177) also being responsible for the models.
Infrared-based approaches have tended to focus on prediction of broad-scale indicators
such as the size of the SMB and C and N mineralization. Fewer studies reported specific
indicators such as enzyme assays, microbial community components (e.g., fungi, bacteria),
or gene quantification (16S RNA). For most of the reported biological properties referred
to in this review, predictions were available for the Vis-NIR and NIR regions, and the MIR
region was also used for the prediction of enzymatic activities (33, 173, 178), microbial
biomass C and N (3, 33, 173, 179), N mineralization (33, 116), and 16S RNA (173).
Although fewer attempts have been reported for prediction of soil biological com-
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

pared to chemical properties, the literature suggests good potential in developing rapid,
cost-effective, infrared-based methodologies to support monitoring and mapping soil bio-
logical parameters. In developing biologically related predictions, sample pretreatment and
measurement data inconsistency are particularly relevant considerations. Sample pretreat-
ment (e.g., sieving, pre-incubation, water content), incubation conditions, kinetic-based
assay measurements, and variation in extraction efficiency between different soil types
can strongly influence the background calibration/validation data set. Mixed prediction
success may relate to the type of soils contained within the calibration/validation data
sets. For example, the soil microbial biomass estimated through the fumigation–extraction
method (180) relies on the use of an extraction efficiency factor. Variation in extraction
efficiency may add noise to the measured data set when the calibration/validation data set
contains a broad range of soils. Local calibrations may avoid extraction efficiency biases
to a greater degree compared to larger global calibration. Terhoeven-Urselmans et al. (160)
also highlighted the importance of considering sample pretreatment, with reduced suc-
cess in predicting biomass-P when samples have undergone pre-incubation compared to
field-moist samples. Consideration of sample pretreatment is of obvious relevance across
all biological measures, where disturbance and pretreatment conditions can have a strong
influence on the size, composition, and activity of the soil microbial community.
The level of microbial diversity, degree of functional redundancy, and high level of
spatial variability at a fine scale make it difficult to develop universal or global models
(14, 33, 181, 182). Furthermore, calibration/validation against similar populations may
provide more robust outcomes. Because soil quality is multifaceted, including consideration
of physical, chemical, and biological parameters, Velasquez et al. (183) suggested the
use of general indicators of soil quality that integrate several biological measurements.
Cécillon et al. (184) demonstrated successful (R2 > 0.9) NIR-based determination of
three such indices that include organic matter, nutrient supply, and broad-scale biological
activity. From an infrared viewpoint, these approaches may offer the greatest potential as an
integrator of spectral information to predict general indicators of soil quality–type indexes
that combine several biological aspects.

Microbial Biomass. Microbial biomass determinations can be useful for a better quantifi-
cation of nutrient cycling in soil (171). The best predictions for biomass C were found
for MIR, NIR, and Vis-NIR, with predictions classified as moderately successful (R2 =
0.82 to 0.84). However, only three attempts to predict this variable using MIR were avail-
able. Visible-NIR was the spectral region that resulted in the highest median R2 values for
biomass N (R2 = 0.93). The general relationship between biomass N and total N is likely
170 J. M. Soriano-Disla et al.

responsible for the good predictions (185). In the case of biomass P, only two studies were
available, with the best results reported in the study of Terhoeven-Urselmans et al. (160)
using Vis-NIR (R2 = 0.81, RPD = 2.0). Overall, Vis-NIR results for microbial biomass N
and P were better than those obtained by using other spectral ranges.
Biomass C/biomass N and biomass C/biomass P PLSR models were calculated in the
study of Terhoeven-Urselmans et al. (160) using Vis-NIR, resulting in R2 values of 0.69 and
0.55 (RPD = 1.6 and 1.3), respectively. Visible-NIR was also used for predicting biomass
C/OC and substrate induced respiration (175, 182), with R2 values of 0.86 and 0.73 (RPD
= 1.5 and 1.7).

Microbial Respiration. When used alone, the size of the SMB has limited value as a
biological indicator (186). The SMB is commonly reported alongside substrate-induced
respiration and basal soil respiration (BSR) rate or used to calculate the metabolic quotient
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

(respiration per unit biomass) as an indicator of the efficiency of C utilization. Substrate-


induced respiration, reflecting the ability of the SMB to respond to added substrate, has
been predicted with moderate (R2 = 0.73) success (175, 182). A number of studies reported
greater prediction success for BSR compared to the size of the SMB-C (96). In general,
greater prediction success was reported for SMB-C and BSR, compared to prediction of the
metabolic quotient (the ratio of the size of the microbial biomass to the rate of respiration)
(100, 160). Zornoza et al. (100) suggested that the efficiency in C use is strongly controlled
by the physiological status of the microbial community, which is influenced by factors that
are unlikely to have strong spectral intensities. For example, pH is known to be a strong
driver of microbial community structure, function, and respiratory stress but is generally
poorly reflected in spectra.
The process of organic matter mineralization is central to crop nutrient management
and C sequestration and is therefore commonly used as an indicator of soil quality (182).
The majority of studies predicting mineralization were performed by using NIR and Vis-
NIR, with the best results obtained using the former (median R2 = 0.82 and 0.79 for C and
N mineralization, respectively).
Similar results were found for C and N mineralization. Fystro (187) found analogous
spectral responses for TOC, total N, and their mineralization parameters, probably due
to intercorrelation between the parameters. Microbial respiration is the result of the min-
eralization of a range of soil organic substances. These organic substances are infrared
active, allowing for the development of calibrations for microbial respiration. Fystro (187)
demonstrated moderate success (R2 = 0.75–0.84) for the prediction of N mineralization po-
tential (aerobic) outperforming other soil chemical indicators (R2 < 0.77). Moderate success
(R2 > 0.79) has also been demonstrated in prediction (NIR and combined Vis-NIR-MIR)
of the anaerobic potentially mineralizable N, an assay considered to reflect mineralization
of organic N pools. Potentially mineralizable N can be used as an indicator with direct
relevance to the supply of N to a crop (188, 189).
Overall, both Vis-NIR and NIR (more evidence is available for the former) can be used
to predict C and N mineralization with moderate usefulness/success.

Microbial Groups. Because the size of the SMB alone is considered a coarse biological
indicator, more detailed measures of the microbial community composition are often sought
in order to understand the impact of disturbance or management on microbial populations.
Microbial biomarkers, such as ergosterol (a component of fungal membranes), phospho-
lipid fatty acids (PLFAs), or fatty acid methyl esters (FAME), have been widely used as
broad-scale indicators of change in community structure, all of which have attracted some
Soil Properties 171

predictive efforts (see Table 2). The literature reports mainly NIR-based studies with a
median R2 of 0.75 (moderately useful predictions).
Moderately useful (R2 = 0.76, RPD = 2.2) and moderately successful (R2 = 0.88,
RPD = 1.9) accuracies were found in the study by Terhoeven-Urselmans et al. (160) for the
prediction of ergosterol (as an indicator of fungal biomass) and its ratio to SMB-C using Vis-
NIR. Using PLFA biomarkers, Zornoza et al. (100) also demonstrated useful predictions of
fungal biomass across a wide range of soils (R2 = 0.77, RPD = 2.1). Their work reported
successful prediction of biomarker groups including total PLFA biomass, total bacterial
indicators, and actinomycete indicators but limited success in other groups (gram-negative
bacteria, protozoa). Rinnan and Rinnan (133) reported moderately successful (R2 > 0.75)
prediction of total PLFAs, fungi, and the ratio of fungi to bacteria using NIR. Similarly, the
prediction of individual fatty acid methyl esters peaks (gas chromatography [GC]-FAME)
using MIR spectra (R2 = 0.50–0.75) (190) suggested that there may be some potential to
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

develop a rapid surrogate indicators of microbial community structure.


Overall, the literature indicates that microbial groups or compounds found in microbes
can be predicted mainly using NIR, with a median R2 of 0.75 (moderately useful). Though
the cost-effectiveness of replacing time-consuming chemical extractions and GC analysis
with spectrally based predictions is obvious, there is debate over the assignment of some
fatty acid biomarkers to specific microbial groups (191), which may limit interpretation.
From a fundamental ecology perspective, the resolution of biomarker microbial profiling
(PLFA, GC-FAME, ergosterol) has been largely superseded by molecular DNA and RNA
methods. Recently, moderate success (R2 = 0.88, RPD = 2.9) was reported for MIR
predictions of 16S rRNA gene copy number (173) and could be extended to include a wider
range of molecular probes (e.g., fungal, archaeal, specific functional groups) to overcome
the type of assignment uncertainties associated with PLFA biomarker approaches.

Enzyme Activities. Organic matter mineralization is dependent on microbial enzyme activ-


ities that are commonly used as indicators of soil quality or the ability of the soil to cycle C
and nutrients (171). Predictive approaches have been applied to predict enzyme activities
(Table 2) with comparable success reported between NIR and Vis-NIR (R2 = 0.80 and 0.78
respectively).
High R2 values (> 0.80 and as high as 0.93) have been reported for the prediction
of enzymatic activities in a number of studies (100, 172–174, 176, 182, 192). However,
relatively poor performance (R2 < 0.7) was obtained for the predictions of arylsulfatase,
dehydrogenase, and urease in the study by Reeves et al. (172); for acid phosphatase,
arylsulfatase, and dehydrogenase in the study by Reeves et al. (33); for acid phosphatase
and arylsulfatase in the study by Mimmo et al. (178); and for peptidase in the study by
Cohen et al. (20). Reeves et al. (172) concluded that infrared approaches may only be
appropriate where high accuracy is not required. The mixed success may relate to the type
of calibration populations (global versus local). For example, the study by Reeves at al.
(33) included soils collected from different depths and where chemical parameters (e.g., C,
N, pH) and the size and activity of the microbial biomass varied considerably. However,
this was not the case of the study by Mimmo et al. (178) where poor results were found
for arylsulfatase and acid phosphatase using samples coming from the first 10 cm of a
single 20-ha field. In their study, the very low values found for the enzymatic activities
are likely to be responsible for the poor predictions. In order to provide a robust database
from which to predict enzyme functions, special attention needs be paid to the accuracy
and appropriateness of the analytical method. Careful consideration should be given to
the assay optimization (e.g., pH), and the influence of substrate inhibition, cooperative
172 J. M. Soriano-Disla et al.

binding, non-Michaelis-Menton kinetic enzyme reactions, and isofunctional enzymes in


environmental samples, as discussed by German et al. (193).
In relation to the spectral basis for the enzyme activity prediction, Reeves et al. (172)
stated that indirect correlations with C or N may only partially explain the NIR results
obtained. Importantly, measures of biologically active N may be the basis for the enzyme
activity determinations.

Conclusions and Recommendations


There is unquestionable potential for using Vis-NIR and MIR for the rapid and cost-effective
prediction of soil physical, chemical, and biological properties in both the laboratory and in
the field, that is, for proximal soil sensing. The techniques can be used in different scientific
disciplines: for example, health risk assessment, environmental remediation, soil fertility,
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

land management, soil amelioration, C sequestration, irrigation, soil forensics, geological


surveys, and soil mapping.
In relation to the questions that we proposed at the outset,

• An extensive range of soil properties can be predicted using reflectance spectroscopy


(Table 2). From the total number of soil properties predicted by soil spectroscopy
(Table 3), 54, 30, and 43% for MIR, NIR, and Vis-NIR, respectively, presented a
median R2 value higher than 0.80 (moderately successful calibrations). Considering
the accuracy of the predictions and the number of examples available, prediction of
only the following attributes can be considered reliable: soil water content, clay, sand,
organic C (fractions), inorganic C, TOC or SOM, total C, C/N, CEC, exchangeable
Ca and Mg, total N, pH, total concentration of potential pollutant metals/metalloids
(As, Cd, Hg and Pb), enzyme activities, biomass C, and C and N mineralization.
• Although successful predictions were obtained for other soil properties, more re-
search is still needed to confirm prediction capabilities for aggregation; total concen-
trations of Ca, K, Mg, P, and S; different N forms (N fractions, mineral, and organic);
P (sorption); micronutrients and other elements; mineralogical composition; and soil
contaminants.
• The most suitable spectral regions for the prediction of soil properties were MIR,
NIR, and Vis-NIR, with much of the literature focused on these three spectral regions.
The extension of spectral ranges to the combinations NIR-MIR, UV-Vis-NIR, and
Vis-NIR-MIR generally resulted in worse or no prediction improvement compared
to single spectral ranges. The few predictions available for soil properties using Vis
or UV have resulted in poor calibrations.
• The best predictions for physical and chemical properties were mostly found for the
MIR spectral range; however, the selection of the best spectral range is dependent
on the soil property. For physical properties, MIR and NIR appeared to be the most
useful regions. In the case of chemical properties, there was a slight advantage with
MIR, closely followed by Vis-NIR and finally NIR. In contrast, the best results for
biological properties were found for Vis-NIR, with fewer studies using MIR, likely
because useful information related to SOC and SMB can be masked by soil minerals
strongly absorbing in the MIR region.
• Much of the effort in developing predictions for soils has been centered on chemical
properties.
• References reporting calibrations using portable devices (mostly Vis-NIR and NIR)
account for only a small percentage (around 10%) of the total number of references
Soil Properties 173

available, so comparisons are still difficult. Despite the presence of water in many
field soils while scanning and variations in particle size of samples presented to the
spectrometer, calibrations are becoming available with reasonably good accuracy
for properties such as clay content, soil water characteristic, TOC, total C/N, and
some contaminants (e.g., TPHs, PAHs).

When developing a calibration based on infrared spectroscopy for the prediction of


soil properties in unknown samples, the subsequent recommendations (organized at the
different calibration stages) should be followed:

• The concentration, spatial variability, and temporal changes of the property of interest
imply a higher sampling effort, often requiring the development of local models.
• Models for soil properties related to the chemistry of the soil matrix are likely
to be more successful (e.g., available P determination may not be dependent on
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

soil composition). Variables that are predicted by virtue of their correlations with
infrared-active soil properties (indirect calibrations) frequently require the develop-
ment of models for specific soil types, locations, and particular environments.
• The same soil preparation protocols and laboratory methods should be used for all
of the samples to be used in a model (e.g., the same TOC and particle size methods)
and work only with high-quality laboratory data (accuracy of spectroscopy models
and predictions are adversely affected by poor or low resolution analytical data; e.g.,
those related with individual microbial biomarkers).
• In relation to the instrumentation and spectra, we have to select the most suitable
instrument for the intended application (considering portability, accuracy, noise
level, stability, spectral information, and range), choose the spectral regions that are
most sensitive to the soil property of interest (e.g., Vis-NIR versus MIR), and, for
some analytes (e.g., mineral N) use more suitable techniques alternative to DRIFT
(e.g., ATR).
• Once we have the spectra of the soil samples we have to correctly match this
information with the analytical data. First of all, we need to correctly interpret the
spectral information relevant to the model (e.g., carbonate peaks if we want to predict
soil pH), remove noisy and/or irrelevant parts of the spectra, and apply relevant data
transformations (e.g., log transform of laboratory data and baseline correction) to
overcome nonlinearity and asymmetric distributions.
• We need to assess the most relevant multivariate modeling methods (e.g., PLSR,
neural networks, support vectors) according to the experience of the user and char-
acteristics of the variable to model and the inclusion in the models of auxiliary
variables that are readily available (e.g., pH, location, crop type and altitude) that
are not well represented in the infrared spectra.
• We have to ensure that the calibration set adequately represents the unknown samples.
• Once developed, models must be regularly updated with the addition of new samples
to account for outliers in relation to soil variability. Finally, the intended application
determines the degree of accuracy required (e.g., different accuracies are required
for C accounting purposes compared to general soil description).

Acknowledgments
The authors gratefully acknowledge the Grains Research and Development Corporation–
University of South Australia (GRDC-USA00012) project for funding this review. José M.
174 J. M. Soriano-Disla et al.

Soriano Disla also thanks the Department of Education (Government of Valencia, Spain)
for a postdoctoral fellowship (APOSTD/2011/034).

References
1. Saxton, K.E., Rawls, W.J., Romberger, J.S., and Papendick, R.I. (1986) Estimating generalized
soil–water characteristics from texture. Soil Sci. Soc. Am. J., 50: 1031–1036.
2. Minasny, B., McBratney, A.B., and Bristow, K.L. (1999) Comparison of different approaches to
the development of pedotransfer functions for water-retention curves. Geoderma, 93: 225–253.
3. Janik, L.J., Merry, R.H., and Skjemstad, J.O. (1998) Can mid infrared diffuse reflectance
analysis replace soil extractions? Aust. J. Exp. Agric., 38: 681–696.
4. Viscarra Rossel, R.A., Adamchuk, V.I., Sudduth, K.A., McKenzie, N.J., and Lobsey, C. (2011)
Proximal soil sensing. An effective approach for soil measurements in space and time. Adv.
Agron., 113: 237–282.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

5. Malley, D.F., Martin, P.D., and Ben-Dor, E. (2004) Application in analysis of soils. In Near-
Infrared Spectroscopy in Agriculture, Agronomy 44. Roberts, C.A., Workman, J., Jr., and
Reeves, J.B., III, Eds. American Society of Agronomy, Crop Science Society of America, Soil
Science Society of America: Madison, WI, pp. 729–784.
6. Cécillon, L., Barthès, B.G., Gomez, C., Ertlen, D., Genot, V., Hedde, M., Stevens, A., and
Brun, J.J. (2009) Assessment and monitoring of soil quality using near-infrared reflectance
spectroscopy (NIRS). Eur. J. Soil Sci., 60: 770–784.
7. Nduwamungu, C., Ziadi, N., Tremblay, G.F., Parent, L.É., Tremblay, G.F., and Thuriès, B.
(2009) Near-infrared reflectance spectroscopy prediction of soil properties: Effects of sample
cups and preparation. Soil Sci. Soc. Am. J., 73: 1896–1903.
8. Stenberg, B., Viscarra Rossel, R.A., Mouazen, A.M., and Wetterlind, J. (2010) Visible and near
infrared spectroscopy in soil science. Adv. Agron., 107: 163–215.
9. Bellon-Maurel, V., Fernandez-Ahumada, E., Palagos, B., Roger, J.M., and McBratney, A.
(2010) Critical review of chemometric indicators commonly used for assessing the quality of
the prediction of soil attributes by NIR spectroscopy. Trends Anal. Chem., 29: 1073–1081.
10. Viscarra Rossel, R.A., Walvoort, D.J.J., McBratney, A.B., Janik, L.J., and Skjemstad, J.O.
(2006) Visible, near infrared, mid infrared or combined diffuse reflectance spectroscopy for
simultaneous assessment of various soil properties. Geoderma, 131: 59–75.
11. Stenberg, B. and Viscarra-Rossel, R.A. (2010) Diffuse reflectance spectroscopy for high-
resolution soil sensing. In Proximal Soil Sensing, Progress in Soil Science 1, Viscarra-Rossel,
R.A., Ed. Springer Science+Business Media: New York, pp. 29–47.
12. Bellon-Maurel, V. and McBratney, A. (2011) Near-infrared (NIR) and mid-infrared (MIR)
spectroscopic techniques for assessing the amount of carbon stock in soils—Critical review and
research perspectives. Soil Biol. Biochem., 43: 1398–1410.
13. Reeves, J.B., III. (2010) Near- versus mid-infrared diffuse reflectance spectroscopy for soil
analysis emphasizing carbon and laboratory versus on-site analysis: Where are we and what
needs to be done? Geoderma, 158: 3–14.
14. Du, C. and Zhou, J. (2009) Evaluation of soil fertility using infrared spectroscopy: A review.
Environ. Chem. Lett., 7: 97–113.
15. Kuang, B., Mahmood, H.S., Quraishi, M.Z., Hoogmoed, W.B., Mouazen, A.M., and van Henten,
E.J. (2012) Sensing soil properties in the laboratory, in situ, and on-line. A review. Adv. Agron.,
114: 155–223.
16. Nguyen, T.T., Janik, L.J., and Raupach, M. (1991) Diffuse reflectance infrared Fourier transform
(DRIFT) spectroscopy in soil studies. Aust. J. Soil Res., 29: 49–67.
17. Reeves, J.B., III, McCarty, G.W., and Meisinger, J.J. (1999) Near infrared reflectance spec-
troscopy for the analysis of agricultural soils. J. Near Infrared Spectrosc., 7: 179–193.
18. Viscarra Rossel, R.A., Chappell, A., De Caritat, P., and Mckenzie, N.J. (2011) On the
soil information content of visible–near infrared reflectance spectra. Eur. J. Soil Sci., 62:
442–453.
Soil Properties 175

19. Joffre, R., Ågren, G.I., Gillon, D., and Bosatta, E. (2001) Organic matter quality in ecological
studies: Theory meets experiment. Oikos, 93: 451–458.
20. Cohen, M.J., Prenger, J.P., and DeBusk, W.F. (2005) Visible–near infrared reflectance spec-
troscopy for rapid, nondestructive assessment of wetland soil quality. J. Environ. Qual., 34:
1422–1434.
21. Velasquez, E., Lavelle, P., Barrios, E., Joffre, R., and Reversat, F. (2005) Evaluating soil quality
in tropical agroecosystems of Colombia using NIRS. Soil Biol. Biochem., 37: 889–898.
22. Vågen, T.G., Shepherd, K.D., and Walsh, M.G. (2006) Sensing landscape level change in soil
fertility following deforestation and conversion in the highlands of Madagascar using Vis-NIR
spectroscopy. Geoderma, 133: 281–294.
23. Omuto, C.T. (2008) Assessment of soil physical degradation in eastern Kenya by use of a
sequential soil testing protocol. Agric. Ecosyst. Environ., 128: 199–211.
24. Viscarra Rossel, R.A. (2011) Fine-resolution multiscale mapping of clay minerals in Australian
soils measured with near infrared spectra. J. Geophys. Res. F: Earth Surf., 116.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

25. Viscarra Rossel, R.A. and Chen, C. (2011) Digitally mapping the information content
of visible–near infrared spectra of surficial Australian soils. Remote Sens. Environ., 115:
1443–1455.
26. Viscarra Rossel, R.A. and Webster, R. (2011) Discrimination of Australian soil horizons and
classes from their visible–near infrared spectra. Eur. J. Soil Sci., 62: 637–647.
27. Forouzangohar, M., Cozzolino, D., Smernik, R.J., Baldock, J.A., Forrester, S.T., Chittleborough,
D.J., and Kookana, R.S. (2013) Using the power of C-13 NMR to interpret infrared spectra of
soil organic matter: A two-dimensional correlation spectroscopy approach. Vib. Spectrosc., 66:
76–82.
28. Viscarra Rossel, R.A., Rizzo, R., Demattê, J.A.M., and Behrens, T. (2010) Spatial modeling of
a soil fertility index using visible–near-infrared spectra and terrain attributes. Soil Sci. Soc. Am.
J., 74: 1293–1300.
29. Van der Marel, H.W. and Beutelspacher, H. (1976) Clay and related minerals. In Atlas of Infrared
Spectroscopy of Clay Minerals and Their Admixtures, Van der Marel, H.W. and Beutelspacher,
H., Eds. Elsevier Scientific: Amsterdam, The Netherlands, p. 396.
30. Coates, J. (2000) Interpretation of infrared spectra, a practical approach. In Encyclopedia of
Analytical Chemistry, Meyers, R.A., Ed. Wiley & Sons Ltd.: Chichester, UK, pp. 10815–10837.
31. Viscarra-Rossel, R.A. and Behrens, T. (2010) Using data mining to model and interpret soil
diffuse reflectance spectra. Geoderma, 158: 46–54.
32. Sherman, D.M. and Waite, T.D. (1985) Electronic spectra of Fe3+ oxides and oxide hydroxides
in the near IR to near UV. Am. Mineral., 70: 1262–1269.
33. Reeves, J.B., III, McCarty, G.W., and Reeves, V.B. (2001) Mid-infrared diffuse reflectance spec-
troscopy for the quantitative analysis of agricultural soils. J. Agric. Food Chem., 49: 766–772.
34. Grinand, C., Barthès, B.G., Brunet, D., Kouakoua, E., Arrouays, D., Jolivet, C., Caria, G., and
Bernoux, M. (2012) Prediction of soil organic and inorganic carbon contents at a national scale
(France) using mid-infrared reflectance spectroscopy (MIRS). Eur. J. Soil Sci., 63: 141–151.
35. Janik, L.J., Skjemstad, J.O., Shepherd, K.D., and Spouncer, L.R. (2007) The prediction of soil
carbon fractions using mid-infrared–partial least square analysis. Aust. J. Soil Res., 45: 73–81.
36. Dardenne, P., Sinnaeve, G., and Baeten, V. (2000) Multivariate calibration and chemometrics
for near infrared spectroscopy: Which method? J. Near Infrared Spectrosc., 8: 229–237.
37. Brown, D.J., Bricklemyer, R.S., and Miller, P.R. (2005) Validation requirements for diffuse
reflectance soil characterization models with a case study of VNIR soil C prediction in Montana.
Geoderma, 129: 251–267.
38. Varmuza, K. and Filmoser, P. (2009) Introduction to Multivariate Statistical Analysis in Chemo-
metrics. Taylor & Francis: Boca Raton, FL.
39. Shepherd, K.D. and Walsh, M.G. (2002) Development of reflectance spectral libraries for
characterization of soil properties. Soil Sci. Soc. Am. J., 66: 988–998.
40. Viscarra Rossel, R.A., Jeon, Y.S., Odeh, I.O.A., and McBratney, A.B. (2008) Using a legacy
soil sample to develop a mid-IR spectral library. Aust. J. Soil Res., 46: 1–16.
176 J. M. Soriano-Disla et al.

41. Brown, D.J., Shepherd, K.D., Walsh, M.G., Dewayne Mays, M., and Reinsch, T.G. (2006) Global
soil characterization with VNIR diffuse reflectance spectroscopy. Geoderma, 132: 273–290.
42. Udelhoven, T., Emmerling, C., and Jarmer, T. (2003) Quantitative analysis of soil chemical
properties with diffuse reflectance spectrometry and partial least-square regression: A feasibility
study. Plant Soil, 251: 319–329.
43. Viscarra Rossel, R.A. and Webster, R. (2012) Predicting soil properties from the Australian soil
visible–near infrared spectroscopic database. Eur. J. Soil Sci., 63: 848–860.
44. Naes, T., Isaksson, T., and Kowalski, B. (1990) Locally weighted regression and scatter correc-
tion for near-infrared reflectance data. Anal. Chem., 62: 664–673.
45. Shenk, J.S., Westerhaus, M.O., and Berzaghi, P. (1997) Investigation of a LOCAL calibration
procedure for near infrared instruments. J. Near Infrared Spectrosc., 5: 223–232.
46. Kennard, R.W. and Stone, L.A. (1969) Computer aided design of experiments. Technometrics,
11: 137–148.
47. Ramirez-Lopez, L., Behrens, T., Schmidt, K., Rossel, R.A.V., Demattê, J.A.M., and Scholten,
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

T. (2013) Distance and similarity-search metrics for use with soil Vis-NIR spectra. Geoderma,
199: 43–53.
48. Reeves, J.B., McCarty, G.W., and Hively, W.D. (2010) Mid-versus near-infrared spectroscopy
for on-site analysis of soil. In Proximal Soil Sensing, Viscarra-Rossel, R.A., McBratney, A.B.,
and Minasny, B., Eds. Springer Science+Business Media: New York, pp. 133–142.
49. Conyers, M.K., Poile, G.J., Oates, A.A., Waters, D., and Chan, K.Y. (2011) Comparison of
three carbon determination methods on naturally occurring substrates and the implication for
the quantification of “soil carbon.” Soil Res., 49: 27–33.
50. Hammes, K., Schmidt, M.W.I., Smernik, R.J., Currie, L.A., Ball, W.P., Nguyen, T.H., Lou-
chouarn, P., Houel, S., Gustafsson, Ö, Elmquist, M., Cornelissen, G., Skjemstad, J.O., Masiello,
C.A., Song, J., Peng, P., Mitra, S., Dunn, J.C., Hatcher, P.G., Hockaday, W.C., Smith, D.M.,
Hartkopf-Fröder, C., Böhmer, A., Lüer, B., Huebert, B.J., Amelung, W., Brodowski, S., Huang,
L., Zhang, W., Gschwend, P.M., Flores-Cervantes, D.X., Largeau, C., Rouzaud, J., Rumpel, C.,
Guggenberger, G., Kaiser, K., Rodionov, A., Gonzalez-Vila, F.J., Gonzalez-Perez, J.S., de la
Rosa, J.M., Manning, D.A.C., López-Capél, E., and Ding, L. (2007) Comparison of quantifi-
cation methods to measure fire-derived (black-elemental) carbon in soils and sediments using
reference materials from soil, water, sediment and the atmosphere. Global Biogeochem. Cycles,
21:
51. Davies, B.E. (1971) A statistical comparison of pH values of some English soils after measure-
ment in both water and 0.01M calcium chloride. Soil Sci. Soc. Am. J., 35: 551–552.
52. Bowman, G. and Hutka, J. (2002) Particle size analysis. In Soil Physical Measurement and
Interpretation for Land Evaluation, McKenzie, N., Coughlan, K., and Cresswell, H., Eds.
CSIRO Publishing: Collingwood, Australia, pp. 224–239.
53. Viscarra Rossel, R.A., Fouad, Y., and Walter, C. (2008) Using a digital camera to measure soil
organic carbon and iron contents. Biosyst. Eng., 100: 149–159.
54. Stumpe, B., Weihermüller, L., and Marschner, B. (2011) Sample preparation and selection
for qualitative and quantitative analyses of soil organic carbon with mid-infrared reflectance
spectroscopy. Eur. J. Soil Sci., 62: 849–862.
55. Chang, C.W., Laird, D.A., and Hurburgh, C.R., Jr. (2005) Influence of soil moisture on near-
infrared reflectance spectroscopic measurement of soil properties. Soil Sci., 170: 244–255.
56. Viscarra Rossel, R.A., Cattle, S.R., Ortega, A., and Fouad, Y. (2009) In situ measurements of
soil colour, mineral composition and clay content by Vis-NIR spectroscopy. Geoderma, 150:
253–266.
57. Stenberg, B. (2010) Effects of soil sample pretreatments and standardised rewetting as interacted
with sand classes on Vis-NIR predictions of clay and soil organic carbon. Geoderma, 158: 15–22.
58. Tekin, Y., Tumsavas, Z., and Mouazen, A.M. (2012) Effect of moisture content on prediction of
organic carbon and pH using visible and near-infrared spectroscopy. Soil Sci. Soc. Am. J., 76:
188–198.
59. Jahn, B.R., Linker, R., Upadhyaya, S.K., Shaviv, A., Slaughter, D.C., and Shmulevich, I. (2006)
Mid-infrared spectroscopic determination of soil nitrate content. Biosyst. Eng., 94: 505–515.
Soil Properties 177

60. Linker, R., Weiner, M., Shmulevich, I., and Shaviv, A. (2006) Nitrate determination in soil
pastes using attenuated total reflectance mid-infrared spectroscopy: Improved accuracy via soil
identification. Biosyst. Eng., 94: 111–118.
61. D’Acqui, L.P., Churchman, G.J., Janik, L.J., Ristori, G.G., and Weissmann, D.A. (1999) Effect
of organic matter removal by low-temperature ashing on dispersion of undisturbed aggregates
from a tropical crusting soil. Geoderma, 93: 311–324.
62. Churchman, G.J., Foster, R.C., D’Acqui, L.P., Janik, L.J., Skjemstad, J.O., Merry, R.H., and
Weissmann, D.A. (2010) Effect of land-use history on the potential for carbon sequestration in
an Alfisol. Soil Tillage Res., 109: 23–35.
63. Savitzky, A. and Golay, M.J.E. (1964) Smoothing and differentiation of data by simplified least
squares procedures. Anal. Chem., 36: 1627–1639.
64. Pichevin, L., Bertrand, P., Boussafir, M., and Disnar, J.R. (2004) Organic matter accumulation
and preservation controls in a deep sea modern environment: An example from Namibian slope
sediments. Org. Geochem., 35: 543–559.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

65. Ibrahim, M., Hameed, A.J., and Jalbout, A. (2008) Molecular spectroscopic study of River Nile
sediment in the greater Cairo region. Appl. Spectrosc., 62: 306–311.
66. Griffiths, P.R. and Fuller, M.P. (1982) Mid-infrared spectrometry of powdered samples. In
Advances in Infrared and Raman Spectroscopy, Vol. 9, Clark, R.J.H. and Hester, R.E., Eds.
Heyden: London, pp. 63–129.
67. Bishop, J.L., Murchie, S.L., Pieters, C.M., and Zent, A.P. (2002) A model for formation of
dust, soil, and rock coatings on Mars: Physical and chemical processes on the Martian surface.
J.Geophys. Res. E: Planets, 107: 7-1–7-17.
68. Haaland, D.M. and Thomas, E.V. (1988) Partial least-squares methods for spectral analyses. 1.
Relation to other quantitative calibration methods and the extraction of qualitative information.
Anal. Chem., 60: 1193–1202.
69. Minasny, B. and McBratney, A.B. (2008) Regression rules as a tool for predicting soil properties
from infrared reflectance spectroscopy. Chemom. Intell. Lab. Syst., 94: 72–79.
70. Viscarra Rossel, R.A. and Lark, R.M. (2009) Improved analysis and modelling of soil diffuse
reflectance spectra using wavelets. Eur. J. Soil Sci., 60: 453–464.
71. Janik, L.J., Forrester, S.T., and Rawson, A. (2009) The prediction of soil chemical and physical
properties from mid-infrared spectroscopy and combined partial least-squares regression and
neural networks (PLS-NN) analysis. Chemom. Intell. Lab. Syst., 97: 179–188.
72. Chakraborty, S., Weindorf, D.C., Zhu, Y., Li, B., Morgan, C.L.S., Ge, Y., and Galbraith, J.
(2012) Spectral reflectance variability from soil physicochemical properties in oil contaminated
soils. Geoderma, 177–178: 80–89.
73. Williams, P.C. (1987) Variables affecting near-infrared reflectance spectroscopy. In Near-
Infrared Technology in the Agricultural and Food Industries, 1st ed., Williams, P.C. and Norris,
K.H., Eds. American Association of Cereal Chemists: St Paul, MN, pp. 143–167.
74. Reeves, J.B., III, and Smith, D.B. (2009) The potential of mid- and near-infrared diffuse
reflectance spectroscopy for determining major- and trace-element concentrations in soils from
a geochemical survey of North America. Appl. Geochem., 24: 1472–1481.
75. Shenk, J.S. and Westerhaus, M.O. (1991) New standardisation and calibration procedures for
NIRS analytical systems. Crop Sci., 31: 1694–1696.
76. Fearn, T. (2001) Standardisation and calibration transfer for near infrared instruments: A review.
J. Near Infrared Spectrosc., 9: 229–244.
77. Andrew, A. and Fearn, T. (2004) Transfer by orthogonal projection: Making near-infrared
calibrations robust to between-instrument variation. Chemom. Intell. Lab. Syst., 72: 51–56.
78. Bogrekci, I. and Lee, W.S. (2007) Comparison of ultraviolet, visible, and near infrared sensing
for soil phosphorus. Biosyst. Eng., 96: 293–299.
79. Nocita, M., Kooistra, L., Bachmann, M., Müller, A., Powell, M., and Weel, S. (2011) Predictions
of soil surface and topsoil organic carbon content through the use of laboratory and field
spectroscopy in the Albany Thicket Biome of Eastern Cape Province of South Africa. Geoderma,
167–168: 295–302.
178 J. M. Soriano-Disla et al.

80. Ludwig, B., Nitschke, R., Terhoeven-Urselmans, T., Michel, K., and Flessa, H. (2008) Use of
mid-infrared spectroscopy in the diffuse-reflectance mode for the prediction of the composition
of organic matter in soil and litter. J. Plant Nutr. Soil Sci., 171: 384–391.
81. D’Acqui, L.P., Pucci, A., and Janik, L.J. (2010) Soil properties prediction of western Mediter-
ranean islands with similar climatic environments by means of mid-infrared diffuse reflectance
spectroscopy. Eur. J. Soil Sci., 61: 865–876.
82. Shao, Y. and He, Y. (2011) Nitrogen, phosphorus, and potassium prediction in soils, using
infrared spectroscopy. Soil Res., 49: 166–172.
83. Daniel, K.W., Tripathi, N.K., and Honda, K. (2003) Artificial neural network analysis of labo-
ratory and in situ spectra for the estimation of macronutrients in soils of Lop Buri (Thailand).
Aust. J. Soil Res., 41: 47–59.
84. Mouazen, A.M., De Baerdemaeker, J., and Ramon, H. (2005) Towards development of on-line
soil moisture content sensor using a fibre-type NIR spectrophotometer. Soil Tillage Res., 80:
171–183.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

85. Waiser, T.H., Morgan, C.L.S., Brown, D.J., and Hallmark, C.T. (2007) In situ characterization
of soil clay content with visible near-infrared diffuse reflectance spectroscopy. Soil Sci. Soc.
Am. J., 71: 389–396.
86. Ben-Dor, E., Heller, D., and Chudnovsky, A. (2008) A novel method of classifying soil profiles
in the field using optical means. Soil Sci. Soc. Am. J., 72: 1113–1123.
87. Christy, C.D. (2008) Real-time measurement of soil attributes using on-the-go near infrared
reflectance spectroscopy. Comput. Electron. Agric., 61: 10–19.
88. Gomez, C., Viscarra Rossel, R.A., and McBratney, A.B. (2008) Soil organic carbon prediction
by hyperspectral remote sensing and field Vis-NIR spectroscopy: An Australian case study.
Geoderma, 146: 403–411.
89. Stevens, A., van Wesemael, B., Bartholomeus, H., Rosillon, D., Tychon, B., and Ben-Dor, E.
(2008) Laboratory, field and airborne spectroscopy for monitoring organic carbon content in
agricultural soils. Geoderma, 144: 395–404.
90. Mouazen, A.M., Maleki, M.R., Cockx, L., Van Meirvenne, M., Van Holm, L.H.J., Merckx, R.,
De Baerdemaeker, J., and Ramon, H. (2009) Optimum three-point linkage set up for improving
the quality of soil spectra and the accuracy of soil phosphorus measured using an on-line visible
and near infrared sensor. Soil Tillage Res., 103: 144–152.
91. Bricklemyer, R.S. and Brown, D.J. (2010) On-the-go VisNIR: Potential and limitations for
mapping soil clay and organic carbon. Comput. Electron. Agric., 70: 209–216.
92. Kusumo, B.H., Hedley, M.J., Tuohy, M.P., Hedley, C.P., and Arnold, G.C. (2010) Prediction of
soil carbon and nitrogen concentrations and pasture root densities from proximally sensed soil
spectral reflectance. In Proximal Soil Sensing, Viscarra-Rossel, R.A., McBratney, A.B., and
Minasny, B., Eds. Springer Science+Business Media: New York, pp. 177–190.
93. Kusumo, B.H., Hedley, M.J., Hedley, C.B., and Tuohy, M.P. (2011) Measuring carbon dynamics
in field soils using soil spectral reflectance: Prediction of maize root density, soil organic carbon
and nitrogen content. Plant Soil, 338: 233–245.
94. Kodaira, M. and Shibusawa, S. (2013) Using a mobile real-time soil visible–near infrared sensor
for high resolution soil property mapping. Geoderma, 199: 64–79.
95. Tisdall, J.M. and Oades, J.M. (1982) Organic matter and water-stable aggregates in soils. J. Soil
Sci., 33: 141–163.
96. Chang, C.W., Laird, D.A., Mausbach, M.J., and Hurburgh C.R., J. (2001) Near-infrared re-
flectance spectroscopy—Principal components regression analyses of soil properties. Soil Sci.
Soc. Am. J., 65: 480–490.
97. Madari, B.E., Reeves, J.B., III, Machado, P.L.O.A., Guimarães, C.M., Torres, E., and McCarty,
G.W. (2006) Mid- and near-infrared spectroscopic assessment of soil compositional parameters
and structural indices in two Ferralsols. Geoderma, 136: 245–259.
98. Slaughter, D.C., Pelletier, M.G., and Upadhyaya, S.K. (2001) Sensing soil moisture using NIR
spectroscopy. Appl. Eng. Agric., 17: 241–247.
99. Fearn, T. (2002) Assessing calibrations: SEP, RPD, RER and R2. NIR News, 13: 12–14.
Soil Properties 179

100. Zornoza, R., Guerrero, C., Mataix-Solera, J., Scow, K.M., Arcenegui, V., and Mataix-Beneyto,
J. (2008) Near infrared spectroscopy for determination of various physical, chemical and bio-
chemical properties in Mediterranean soils. Soil Biol. Biochem., 40: 1923–1930.
101. Bertrand, I., Janik, L.J., Holloway, R.E., Armstrong, R.D., and McLaughlin, M.J. (2002) The
rapid assessment of concentrations and solid phase associations of macro- and micronutrients
in alkaline soils by mid-infrared diffuse reflectance spectroscopy. Aust. J. Soil Res., 40: 1339–
1356.
102. Janik, L.J. and Skjemstad, J.O. (1995) Characterization and analysis of soils using mid-infrared
partial least-squares. II. Correlations with some laboratory data. Aust. J. Soil Res., 33: 637–
650.
103. Genot, V., Colinet, G., Bock, L., Vanvyve, D., Reusen, Y., and Dardenne, P. (2011) Near
infrared reflectance spectroscopy for estimating soil characteristics valuable in the diagnosis of
soil fertility. J. Near Infrared Spectrosc., 19: 117–138.
104. Ben-Dor, E. and Banin, A. (1995) Near-infrared analysis as a rapid method to simultaneously
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

evaluate several soil properties. Soil Sci. Soc. Am. J., 59: 364–372.
105. Merry, R.H. and Janik, L.J. (2001) Mid infrared spectroscopy for rapid and cheap analysis of
soils. 10th Australian Agronomy Conference, Hobart, Australia, January 29–February 1.
106. Rayment, G.E. and Lyons, D.J. (2011) Soil Chemical Methods–Australasia. CSIRO Publishing:
Collingwood, Australia.
107. Malley, D.F., Yesmin, L., and Eilers, R.G. (2002) Rapid analysis of hog manure and manure-
amended soils using near-infrared spectroscopy. Soil Sci. Soc. Am. J., 66: 1677–1686.
108. McCarty, G.W., Reeves, J.B., III, Reeves, V.B., Follett, R.F., and Kimble, J.M. (2002) Mid-
infrared and near-infrared diffuse reflectance spectroscopy for soil carbon measurement. Soil
Sci. Soc. Am. J., 66: 640–646.
109. Janik, L.J., Skjemstand, J.O., and Raven, M.D. (1995) Characterization and analysis of soils
using mid-infrared partial least squares. I. Correlations with XRF-determined major element
composition. Aust. J. Soil Res., 33: 621–636.
110. Pirie, A., Singh, B., and Islam, K. (2005) Ultra-violet, visible, near-infrared, and mid-infrared
diffuse reflectance spectroscopic techniques to predict several soil properties. Aust. J. Soil Res.,
43: 713–721.
111. Mouazen, A.M., De Baerdemaeker, J., and Ramon, H. (2006) Effect of wavelength range
on the measurement accuracy of some selected soil constituents using visual–near infrared
spectroscopy. J. Near Infrared Spectrosc., 14: 189–199.
112. Mouazen, A.M., Maleki, M.R., De Baerdemaeker, J., and Ramon, H. (2007) On-line measure-
ment of some selected soil properties using a VIS-NIR sensor. Soil Tillage Res., 93: 13–27.
113. Slade, R., Quirk, J.R., and Norrish, K. (1991) Crystalline swelling of smectite samples in
concentrated NaC1 solutions in relation to layer charge. Clays Clay Miner., 39: 234–238.
114. Laird, D.A. (1999) Layer charge influences on the hydration of expandable 2:1 phyllosilicates.
Clays Clay Miner., 47: 630–636.
115. Slade, P.G. and Quirk, J.P. (1991) The limited crystalline swelling of smectites in CaCl2 , MgCl2 ,
and LaCl3 solutions. J. Colloid Interface Sci., 144: 18–26.
116. Van Groenigen, J.W., Mutters, C.S., Horwath, W.R., and Van Kessel, C. (2003) NIR and DRIFT-
MIR spectrometry of soils for predicting soil and crop parameters in a flooded field. Plant Soil,
250: 155–165.
117. McBratney, A.B., Minasny, B., and Viscarra Rossel, R. (2006) Spectral soil analysis and
inference systems: A powerful combination for solving the soil data crisis. Geoderma, 136:
272–278.
118. Islam, K., Singh, B., and McBratney, A. (2003) Simultaneous estimation of several soil prop-
erties by ultra-violet, visible, and near-infrared reflectance spectroscopy. Aust. J. Soil Res., 41:
1101–1114.
119. Shepherd, K.D. and Walsh, M.G. (2007) Review: Infrared spectroscopy—Enabling an evidence-
based diagnostic surveillance approach to agricultural and environmental management in de-
veloping countries. J. Near Infrared Spectrosc., 15: 1–19.
180 J. M. Soriano-Disla et al.

120. Kim, H.J., Sudduth, K.A., and Hummel, J.W. (2009) Soil macronutrient sensing for precision
agriculture. J. Environ. Monitor., 11: 1810–1824.
121. Bogrekci, I. and Lee, W.S. (2005) Spectral soil signatures and sensing phosphorus. Biosyst.
Eng., 92: 527–533.
122. Soriano-Disla, J.M., Janik, L., McLaughlin, M.J., Forrester, S., Kirby, J., Reimann, C., Albanese,
S., Andersson, M., Arnoldussen, A., Baritz, R., Batista, M.J., Bel-Lan, A., Birke, M., Cicchella,
D., Demetriades, A., Dinelli, E., De Vivo, B., De Vos, W., Dohrmann, R., Duris, M., Dusza-
Dobek, A., Eggen, O.A., Eklund, M., Ernstsen, V., Filzmoser, P., Finne, T.E., Flight, D., Fuchs,
M., Fugedi, U., Gilucis, A., Gosar, M., Gregorauskiene, V., Gulan, A., Halamid, J., Haslinger, E.,
Hayoz, P., Hobiger, G., Hoffmann, R., Hoogewerff, J., Hrvatovic, H., Husnjak, S., Johnson, C.C.,
Jordan, G., Kivisilla, J., Klos, V., Krone, F., Kwecko, P., Kuti, L., Ladenberger, A., Lima, A.,
Locutura, J., Lucivjansky, P., Mackovych, D., Malyuk, B.I., Maquil, R., Meuli, R.G., Miosic, N.,
Mol, G., Négrel, P., O’Connor, P., Oorts, K., Ottesen, R.T., Pasieczna, A., Petersell, V., Pfleiderer,
S., Ponavicc, M., Prazeres, C., Rauch, U., Salpeteur, I., Schedl, A., Scheib, A., Schoeters, I.,
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Sefcik, P., Sellersjö, E., Skopljak, F., Slaninka, I., Šorša, A., Srvkota, R., Stafilov, T., Tarvainen,
T., Trendavilov, V., Valera, P., Verougstraete, V., Vidojevid, D., Zissimos, A.M., and Zomeni,
Z. (2013) The use of diffuse reflectance mid-infrared spectroscopy for the prediction of the
concentration of chemical elements estimated by X-ray fluorescence in agricultural and grazing
European soils. Appl. Geochem., 29: 135–143.
123. Sinfield, J.V., Fagerman, D., and Colic, O. (2010) Evaluation of sensing technologies for on-
the-go detection of macro-nutrients in cultivated soils. Comput. Electron. Agric., 70: 1–18.
124. Öborn, I., Andrist-Rangel, Y., Askekaard, M., Grant, C.A., Watson, C.A., and Edwards, A.C.
(2005) Critical aspects of potassium management in agricultural systems. Soil Use Manage.,
21: 102–112.
125. Bramley, R.G.V. and Janik, L.J. (2005) Precision agriculture demands a new approach to soil
and plant sampling and analysis—Examples from Australia. Commun. Soil Sci. Plant Anal.,
36: 9–22.
126. Whitehead, D.C. (2000) Nutrient Elements in Grassland: Soil–Plant–Animal Relationships.
CAB International: Wallingford, UK.
127. Minasny, B., Tranter, G., McBratney, A.B., Brough, D.M., and Murphy, B.W. (2009) Regional
transferability of mid-infrared diffuse reflectance spectroscopic prediction for soil chemical
properties. Geoderma, 153: 155–162.
128. Van Vuuren, J.A.J., Meyer, J.H., and Claassens, A.S. (2006) Potential use of near infrared
reflectance monitoring in precision agriculture. Commun. Soil Sci. Plant Anal., 37: 2171–2184.
129. Cozzolino, D. and Morón, A. (2003) The potential of near-infrared reflectance spectroscopy to
analyse soil chemical and physical characteristics. J. Agric. Sci., 140: 65–71.
130. Fox, R.H., Shenk, J.S., Piekielek, W.P., Westerhaus, M.O., Toth, J.D., and Macneal, K.E. (1993)
Comparison of near-infrared spectroscopy and other soil nitrogen availability quick tests for
corn. Agron. J., 85: 1049–1053.
131. St. Luce, M., Ziadi, N., Nyiraneza, J., Tremblay, G.F., Zebarth, B.J., Whalen, J.K., and Laterrière,
M. (2012) Near infrared reflectance spectroscopy prediction of soil nitrogen supply in humid
temperate regions of Canada. Soil Sci. Soc. Am. J., 76: 1454–1461.
132. Martin, P.D., Malley, D.F., Manning, G., and Fuller, L. (2002) Determination of soil organic
carbon and nitrogen at the field level using near-infrared spectroscopy. Can. J. Soil Sci., 82:
413–422.
133. Rinnan, R. and Rinnan, A. (2007) Application of near infrared reflectance (NIR) and fluores-
cence spectroscopy to analysis of microbiological and chemical properties of arctic soil. Soil
Biol. Biochem., 39: 1664–1673.
134. Du, C., Zhou, J., Wang, H., Chen, X., Zhu, A., and Zhang, J. (2009) Determination of soil
properties using Fourier transform mid-infrared photoacoustic spectroscopy. Vib. Spectrosc.,
49: 32–37.
135. Ehsani, M.R., Upadhyaya, S.K., Slaughter, D., Shafii, S., and Pelletier, M. (1999) A NIR
technique for rapid determination of soil mineral nitrogen. Precis. Agric., 1: 217–234.
Soil Properties 181

136. Freschet, G.T., Barthès, B.G., Brunet, D., Hien, E., and Masse, D. (2011) Use of near infrared re-
flectance spectroscopy (NIRS) for predicting soil fertility and historical management. Commun.
Soil Sci. Plant Anal., 42: 1692–1705.
137. Maleki, M.R., Van Holm, L., Ramon, H., Merckx, R., De Baerdemaeker, J., and Mouazen, A.M.
(2006) Phosphorus sensing for fresh soils using visible and near infrared spectroscopy. Biosyst.
Eng., 95: 425–436.
138. Burkitt, L.L., Moody, P.W., Gourley, C.J.P., and Hannah, M.C. (2002) A simple phosphorus
buffering index for Australian soils. Aust. J. Soil Res., 40: 497–513.
139. Mason, S., McNeill, A., McLaughlin, M.J., and Zhang, H. (2010) Prediction of wheat response
to an application of phosphorus under field conditions using diffusive gradients in thin-films
(DGT) and extraction methods. Plant Soil, 337: 243–258.
140. Cohen, M.J., Paris, J., and Clark, M.W. (2007) P-sorption capacity estimation in Southeastern
USA wetland soils using visible/near infrared (VNIR) reflectance spectroscopy. Wetlands, 27:
1098–1111.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

141. Morón, A. and Cozzolino, D. (2007) Measurement of phosphorus in soils by near infrared
reflectance spectroscopy: Effect of reference method on calibration. Commun. Soil Sci. Plant
Anal., 38: 1965–1974.
142. He, Y., Huang, M., Garcı́a, A., Hernández, A., and Song, H. (2007) Prediction of soil macronu-
trients content using near-infrared spectroscopy. Comput. Electron. Agric., 58: 144–153.
143. Kizewski, F., Liu, Y.T., Morris, A., and Hesterberg, D. (2011) Spectroscopic approaches for
phosphorus speciation in soils and other environmental systems. J. Environ. Qual., 40: 751–766.
144. Kemper, T. and Sommer, S. (2002) Estimate of heavy metal contamination in soils after a
mining accident using reflectance spectroscopy. Environ. Sci. Technol., 36: 2742–2747.
145. Chodak, M., Ludwig, B., Khanna, P., and Beese, F. (2002) Use of near infrared spectroscopy
to determine biological and chemical characteristics of organic layers under spruce and beech
stands. J. Plant Nutr. Soil Sci., 165: 27–33.
146. Wu, Y., Chen, J., Wu, X., Tian, Q., Ji, J., and Qin, Z. (2005) Possibilities of reflectance
spectroscopy for the assessment of contaminant elements in suburban soils. Appl. Geochem.,
20: 1051–1059.
147. Wu, Y., Chen, J., Ji, J., Gong, P., Liao, Q., Tian, Q., and Ma, H. (2007) A mechanism study
of reflectance spectroscopy for investigating heavy metals in soils. Soil Sci. Soc. Am. J., 71:
918–926.
148. Malley, D.F. and Williams, P.C. (1997) Use of near-infrared reflectance spectroscopy in predic-
tion of heavy metals in freshwater sediment by their association with organic matter. Environ.
Sci. Technol., 31: 3461–3467.
149. Kooistra, L., Wehrens, R., Leuven, R.S.E.W., and Buydens, L.M.C. (2001) Possibilities of
visible–near-infrared spectroscopy for the assessment of soil contamination in river floodplains.
Anal. Chim. Acta, 446: 97–105.
150. Siebielec, G., McCarty, G.W., Stuczynski, T.I., and Reeves, J.B. III. (2004) Near- and mid-
infrared diffuse reflectance spectroscopy for measuring soil metal content. J. Environ. Qual.,
33: 2056–2069.
151. Halliwell, D.J., Barlow, K.M., and Nash, D.M. (2001) A review of the effects of wastewater
sodium on soil physical properties and their implications for irrigation systems. Aust. J. Soil
Res., 39: 1259–1267.
152. Panno, S.V., Hackley, K.C., Hwang, H.H., Greenberg, S.E., Krapac, I.G., Landsberger, S., and
O’Kelly, D.J. (2006) Characterization and identification of Na-Cl sources in ground water.
Ground Water, 44: 176–187.
153. Nag, A., Chakraborty, D., and Chandra, A. (2008) Effects of ion concentration on the hydrogen
bonded structure of water in the vicinity of ions in aqueous NaCl solutions. J. Chem. Sci., 120:
71–77.
154. Abdi, D., Tremblay, G.F., Ziadi, N., Bélanger, G., and Parent, L.E. (2012) Predicting soil
phosphorus-related properties using near-infrared reflectance spectroscopy. Soil Sci. Soc. Am.
J., 76: 2318–2326.
182 J. M. Soriano-Disla et al.

155. Ben-Dor, E. and Banin, A. (1994) Visible and near-infrared (0.4–1.1 μm) analysis of arid and
semiarid soils. Remote Sens. Environ., 48: 261–274.
156. Vendrame, P.R.S., Marchão, R.L., Brunet, D., and Becquer, T. (2012) The potential of NIR
spectroscopy to predict soil texture and mineralogy in Cerrado Latosols. Eur. J. Soil Sci., 63:
743–753.
157. Summers, D., Lewis, M., Ostendorf, B., and Chittleborough, D. (2011) Visible near-infrared
reflectance spectroscopy as a predictive indicator of soil properties. Ecol. Ind., 11: 123–131.
158. Viscarra Rossel, R.A., Bui, E.N., De Caritat, P., and McKenzie, N.J. (2010) Mapping iron oxides
and the color of Australian soil using visible–near-infrared reflectance spectra. J. Geophys. Res.
F: Earth Surf., 115.
159. Sudduth, K.A., Drummond, S.T., and Kitchen, N.R. (2001) Accuracy issues in electromagnetic
induction sensing of soil electrical conductivity for precision agriculture. Comput. Electron.
Agric., 31: 239–264.
160. Terhoeven-Urselmans, T., Schmidt, H., Georg Joergensen, R., and Ludwig, B. (2008) Usefulness
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

of near-infrared spectroscopy to determine biological and chemical soil properties: Importance


of sample pre-treatment. Soil Biol. Biochem., 40: 1178–1188.
161. Kookana, R.S., Janik, L.J., Forouzangohar, M., and Forrester, S.T. (2008) Prediction of atrazine
sorption coefficients in soils using mid-infrared spectroscopy and partial least-squares analysis.
J. Agric. Food Chem., 56: 3208–3213.
162. Singh, B., Malley, D.F., Farenhorst, A., and Williams, P. (2012) Feasibility of using near-
infrared spectroscopy for rapid quantification of 17β-estradiol sorption coefficients in soil. J.
Agric. Food Chem., 60: 9948–9953.
163. Forouzangohar, M., Cozzolino, D., Kookana, R.S., Smernik, R.J., Forrester, S.T., and Chittle-
borough, D.J. (2009) Direct comparison between visible near- and mid-infrared spectroscopy
for describing diuron sorption in soils. Environ. Sci. Technol., 43: 4049–4055.
164. Bray, J.G.P., Rossel, R.V., and Mcbratney, A.B. (2009) Diagnostic screening of urban soil
contaminants using diffuse reflectance spectroscopy. Aust. J. Soil Res., 47: 433–442.
165. Janik, L., Loibner, A.P., Kattner, J., and Edelmann, E. (2011) Method for determining poly-
cyclic aromatic hydrocarbons contaminant concentration. European Patent Office: Rijswijk,
The Netherlands. WO 2011/051166 A1.
166. Malley, D.F., Hunter, K.N., and Webster, G.R.B. (1999) Analysis of diesel fuel contamina-
tion in soils by near-infrared reflectance spectrometry and solid phase microextraction–gas
chromatography. J. Soil Contam., 8: 481–489.
167. Chakraborty, S., Weindorf, D.C., Morgan, C.L.S., Ge, Y., Galbraith, J.M., Li, B., and Kahlon,
C.S. (2010) Rapid identification of oil-contaminated soils using visible near-infrared diffuse
reflectance spectroscopy. J. Environ. Qual., 39: 1378–1387.
168. Chakraborty, S., Weindorf, D.C., Zhu, Y., Li, B., Morgan, C.L.S., Ge, Y., and Galbraith, J.
(2012) Assessing spatial variability of soil petroleum contamination using visible near-infrared
diffuse reflectance spectroscopy. J. Environ. Monitor., 14: 2886–2892.
169. Forrester, S., Janik, L., and McLaughlin, M. (2010) In-situ determination of total petroleum
hydrocarbon (TPH) contamination: A quick infrared spectroscopic test for TPH at contaminated
sites. 19th World Congress of Soil Science. Brisbane, Australia, August 1–6.
170. Forrester, S.T., Janik, L.J., McLaughlin, M.J., Soriano-Disla, J.M., Stewart, R., and Dearman, B.
(2013) Total petroleum hydrocarbon concentration prediction in soils using diffuse reflectance
infrared spectroscopy. Soil Sci. Soc. Am. J., 77: 450–460.
171. Nannipieri, P., Ascher, J., Ceccherini, M.T., Landi, L., Pietramellara, G., and Renella, G. (2003)
Microbial diversity and soil functions. Eur. J. Soil Sci., 54: 655–670.
172. Reeves, J.B., III, McCarty, G.W., and Meisinger, J.J. (2000) Near infrared reflectance spec-
troscopy for the determination of biological activity in agricultural soils. J. Near Infrared
Spectrosc., 8: 161–170.
173. Rasche, F., Marhan, S., Berner, D., Keil, D., Kandeler, E., and Cadisch, G. (2013) MidDRIFTS-
based partial least square regression analysis allows predicting microbial biomass, enzyme
activities and 16S rRNA gene abundance in soils of temperate grasslands. Soil Biol. Biochem.,
57: 504–512.
Soil Properties 183

174. Chodak, M. (2011) Near-infrared spectroscopy for rapid estimation of microbial properties in
reclaimed mine soils. J. Plant Nutr. Soil Sci., 174: 702–709.
175. Chodak, M., Niklińska, M., and Beese, F. (2007) Near-infrared spectroscopy for analy-
sis of chemical and microbiological properties of forest soil organic horizons in a heavy-
metal–polluted area. Biol. Fertil. Soils, 44: 171–180.
176. Cécillon, L., Cassagne, N., Czarnes, S., Gros, R., and Brun, J.J. (2008) Variable selection in
near infrared spectra for the biological characterization of soil and earthworm casts. Soil Biol.
Biochem., 40: 1975–1979.
177. Vaidyanathan, S., Macaloney, G., and McNeil, B. (1999) Fundamental investigations on the
near-infrared spectra of microbial biomass as applicable to bioprocess monitoring. Analyst, 124:
157–162.
178. Mimmo, T., Reeves, J.B., III, McCarty, G.W., and Galletti, G. (2002) Determination of biological
measures by mid-infrared diffuse reflectance spectroscopy in soils within a landscape. Soil Sci.,
167: 281–287.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

179. Reeves, J.B., III, Follett, R.F., McCarty, G.W., and Kimble, J.M. (2006) Can near or mid-
infrared diffuse reflectance spectroscopy be used to determine soil carbon pools? Commun. Soil
Sci. Plant Anal., 37: 2307–2325.
180. Vance, E.D., Brookes, P.C., and Jenkinson, D.S. (1987) An extraction method for measuring
soil microbial biomass C. Soil Biol. Biochem., 19: 703–707.
181. Terhoeven-Urselmans, T., Michel, K., Helfrich, M., Flessa, H., and Ludwig, B. (2006) Near-
infrared spectroscopy can predict the composition of organic matter in soil and litter. J. Plant
Nutr. Soil Sci., 169: 168–174.
182. Schimann, H., Joffre, R., Roggy, J.C., Lensi, R., and Domenach, A.M. (2007) Evaluation of the
recovery of microbial functions during soil restoration using near-infrared spectroscopy. Appl.
Soil Ecol., 37: 223–232.
183. Velasquez, E., Lavelle, P., and Andrade, M. (2007) GISQ, a multifunctional indicator of soil
quality. Soil Biol. Biochem., 39: 3066–3080.
184. Cécillon, L., Cassagne, N., Czarnes, S., Gros, R., Vennetier, M., and Brun, J.J. (2009) Predicting
soil quality indices with near infrared analysis in a wildfire chronosequence. Sci. Total Environ.,
407: 1200–1205.
185. Reeves, J.B. III, and McCarty, G.W. (2001) Quantitative analysis of agricultural soils using near
infrared reflectance spectroscopy and a fibre-optic probe. J. Near Infrared Spectrosc., 9: 25–34.
186. Broos, K., Macdonald, L.M., St. Warne, M.J., Heemsbergen, D.A., Barnes, M.B., Bell, M., and
McLaughlin, M.J. (2007) Limitations of soil microbial biomass carbon as an indicator of soil
pollution in the field. Soil Biol. Biochem., 39: 2693–2695.
187. Fystro, G. (2002) The prediction of C and N content and their potential mineralisation in
heterogeneous soil samples using Vis-NIR spectroscopy and comparative methods. Plant Soil,
246: 139–149.
188. Russell, C.A., Angus, J.F., Batten, G.D., Dunn, B.W., and Williams, R.L. (2002) The potential
of NIR spectroscopy to predict nitrogen mineralization in rice soils. Plant Soil, 247: 243–
252.
189. Murphy, D.V., Osman, M., Russell, C.A., Darmawanto, S., and Hoyle, F.C. (2009) Potentially
mineralisable nitrogen: Relationship to crop production and spatial mapping using infrared
reflectance spectroscopy. Aust. J. Soil Res., 47: 737–741.
190. Pankhurst, C.E., Janik, L.J., Kirkby, C.A., and Hawke, B.G. (1997) Application of GC-FAME
and mid-infrared analysis of soil as a measure of soil health. RIRDC Project CSO-10A, Final
Report. CSIRO Land & Water: Adelaide, Australia.
191. Ngosong, C., Gabriel, E., and Ruess, L. (2012) Use of the signature fatty acid 16:1w5 as a tool
to determine the distribution of arbuscular mycorrhizal fungi in soil. J. Lipids, 2012: 1–8.
192. Dick, W.A., Thavamani, B., Conley, S., Blaisdell, R., and Sengupta, A. (2013) Prediction of
ß-glucosidase and ß-glucosaminidase activities, soil organic C, and amino sugar N in a diverse
population of soils using near infrared reflectance spectroscopy. Soil Biol. Biochem., 56: 99–
104.
184 J. M. Soriano-Disla et al.

193. German, D.P., Weintraub, M.N., Grandy, A.S., Lauber, C.L., Rinkes, Z.L., and Allison, S.D.
(2011) Optimization of hydrolytic and oxidative enzyme methods for ecosystem studies. Soil
Biol. Biochem., 43: 1387–1397.
194. Dalal, R.C. and Henry, R.J. (1986) Simultaneous determination of moisture, organic carbon,
and total nitrogen by near infrared reflectance spectrophotometry. Soil Sci. Soc. Am. J., 50:
120–123.
195. Morra, M.J., Hall, M.H., and Freeborn, L.L. (1991) Carbon and nitrogen analysis of soil fractions
using near-infrared reflectance spectroscopy. Soil Sci. Soc. Am. J., 55: 288–291.
196. Hummel, J.W., Sudduth, K.A., and Hollinger, S.E. (2001) Soil moisture and organic matter
prediction of surface and subsurface soils using an NIR soil sensor. Comput. Electron. Agric.,
32: 149–165.
197. Mouazen, A.M., Saeys, W., Xing, J., De Baerdemaeker, J., and Ramon, H. (2005) Near infrared
spectroscopy for agricultural materials: An instrument comparison. J. Near Infrared Spectrosc.,
13: 87–97.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

198. Mouazen, A.M., Karoui, R., De Baerdemaeker, J., and Ramon, H. (2006) Characterization of
soil water content using measured visible and near infrared spectra. Soil Sci. Soc. Am. J., 70:
1295–1302.
199. Janik, L.J., Merry, R.H., Forrester, S.T., Lanyon, D.M., and Rawson, A. (2007) Rapid prediction
of soil water retention using mid infrared spectroscopy. Soil Sci. Soc. Am. J., 71: 507–514.
200. Sørensen, L.K. and Dalsgaard, S. (2005) Determination of clay and other soil properties by near
infrared spectroscopy. Soil Sci. Soc. Am. J., 69: 159–167.
201. McCarty, G.W. and Reeves, J.B. III. (2006) Comparison of near infrared and mid infrared
diffuse reflectance spectroscopy for field-scale measurement of soil fertility parameters. Soil
Sci., 171: 94–102.
202. Zimmermann, M., Leifeld, J., and Fuhrer, J. (2007) Quantifying soil organic carbon fractions
by infrared-spectroscopy. Soil Biol. Biochem., 39: 224–231.
203. Cobo, J.G., Dercon, G., Yekeye, T., Chapungu, L., Kadzere, C., Murwira, A., Delve, R., and
Cadisch, G. (2010) Integration of mid-infrared spectroscopy and geostatistics in the assessment
of soil spatial variability at landscape level. Geoderma, 158: 398–411.
204. Sudduth, K.A., Kitchen, N.R., Sadler, E.J., Drummond, S.T., and Myers, D.B. (2010) VNIR
spectroscopy estimates of within-field variability in soil properties. In Proximal Soil Sensing,
Viscarra-Rossel, R.A., McBratney, A.B., and Minasny, B., Eds. Springer Science+Business
Media: New York, pp. 153–163.
205. Wetterlind, J., Stenberg, B., and Söderström, M. (2010) Increased sample point density in farm
soil mapping by local calibration of visible and near infrared prediction models. Geoderma,
156: 152–160.
206. Leone, A.P., Viscarra-Rossel, R.A., Amenta, P., and Buondonno, A. (2012) Prediction of soil
properties with PLSR and Vis-NIR spectroscopy: Application to mediterranean soils from
southern Italy. Curr. Anal. Chem., 8: 283–299.
207. Masserschmidt, I., Cuelbas, C.J., Poppi, R.J., De Andrade, J.C., De Abreu, C.A., and Da-
vanzo, C.U. (1999) Determination of organic matter in soils by FTIR/diffuse reflectance and
multivariate calibration. J. Chemom., 13: 265–273.
208. Confalonieri, M., Fornasier, F., Ursino, A., Boccardi, F., Pintus, B., and Odoardi, M. (2001) The
potential of near infrared reflectance spectroscopy as a tool for the chemical characterisation of
agricultural soils. J. Near Infrared Spectrosc., 9: 123–131.
209. Chang, C. and Laird, D.A. (2002) Near-infrared reflectance spectroscopic analysis of soil C and
N. Soil Sci., 167: 110–116.
210. Dunn, B.W., Beecher, H.G., Batten, G.D., and Ciavarella, S. (2002) The potential of near-
infrared reflectance spectroscopy for soil analysis—A case study from the Riverine Plain of
south-eastern Australia. Aust. J. Exp. Agric., 42: 607–614.
211. Fidêncio, P.H., Poppi, R.J., and De Andrade, J.C. (2002) Determination of organic matter in
soils using radial basis function networks and near infrared spectroscopy. Anal. Chim. Acta,
453: 125–134.
Soil Properties 185

212. Moron, A. and Cozzolino, D. (2004) Determination of potentially mineralizable nitrogen and
nitrogen in particulate organic matter fractions in soil by visible and near-infrared reflectance
spectroscopy. J. Agric. Sci., 142: 335–343.
213. Van Waes, C., Mestdagh, I., Lootens, P., and Carlier, L. (2005) Possibilities of near infrared
reflectance spectroscopy for the prediction of organic carbon concentrations in grassland soils.
J. Agric. Sci., 143: 487–492.
214. Butkute, B. and Šlepetiene, A. (2006) Application of near infrared reflectance spectroscopy
for the assessment of soil quality in a long-term pasture. Commun. Soil Sci. Plant Anal., 37:
2389–2409.
215. Leifeld, J. (2006) Application of diffuse reflectance FT-IR spectroscopy and partial least-
squares regression to predict NMR properties of soil organic matter. Eur. J. Soil Sci., 57: 846–
857.
216. Mutuo, P.K., Shepherd, K.D., Albrecht, A., and Cadisch, G. (2006) Prediction of carbon miner-
alization rates from different soil physical fractions using diffuse reflectance spectroscopy. Soil
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

Biol. Biochem., 38: 1658–1664.


217. Brunet, D., Barthès, B.G., Chotte, J., and Feller, C. (2007) Determination of carbon and nitrogen
contents in Alfisols, Oxisols and Ultisols from Africa and Brazil using NIRS analysis: Effects
of sample grinding and set heterogeneity. Geoderma, 139: 106–117.
218. Barthès, B.G., Brunet, D., Hien, E., Enjalric, F., Conche, S., Freschet, G.T., D’Annunzio, R.,
and Toucet-Louri, J. (2008) Determining the distributions of soil carbon and nitrogen in particle
size fractions using near-infrared reflectance spectrum of bulk soil samples. Soil Biol. Biochem.,
40: 1533–1537.
219. Brunet, D., Bernoux, M., and Barthès, B.G. (2008) Comparison between predictions of C
and N contents in tropical soils using a Vis-NIR spectrometer including a fibre-optic probe
versus a NIR spectrometer including a sample transport module. Biosyst. Eng., 100: 448–
452.
220. Morgan, C.L.S., Waiser, T.H., Brown, D.J., and Hallmark, C.T. (2009) Simulated in situ char-
acterization of soil organic and inorganic carbon with visible near-infrared diffuse reflectance
spectroscopy. Geoderma, 151: 249–256.
221. Moros, J., Martı́nez-Sánchez, M.J., Pérez-Sirvent, C., Garrigues, S., and de la Guardia, M.
(2009) Testing of the region of Murcia soils by near infrared diffuse reflectance spectroscopy
and chemometrics. Talanta, 78: 388–398.
222. Fontán, J.M., Calvache, S., López-Bellido, R.J., and López-Bellido, L. (2010) Soil carbon
measurement in clods and sieved samples in a Mediterranean vertisol by visible and near-
infrared reflectance spectroscopy. Geoderma, 156: 93–98.
223. Mouazen, A.M., Kuang, B., De Baerdemaeker, J., and Ramon, H. (2010) Comparison among
principal component, partial least squares and back propagation neural network analyses for
accuracy of measurement of selected soil properties with visible and near infrared spectroscopy.
Geoderma, 158: 23–31.
224. Kamau-Rewe, M., Rasche, F., Cobo, J.G., Dercon, G., Shepherd, K.D., and Cadisch, G. (2011)
Generic prediction of soil organic carbon in alfisols using diffuse reflectance Fourier-transform
mid-infrared spectroscopy. Soil Sci. Soc. Am. J., 75: 2358–2360.
225. Kuang, B. and Mouazen, A.M. (2011) Calibration of visible and near infrared spectroscopy for
soil analysis at the field scale on three European farms. Eur. J. Soil Sci., 62: 629–636.
226. Sarkhot, D.V., Grunwald, S., Ge, Y., and Morgan, C.L.S. (2011) Comparison and detection
of total and available soil carbon fractions using visible/near infrared diffuse reflectance spec-
troscopy. Geoderma, 164: 22–32.
227. Vohland, M., Besold, J., Hill, J., and Fründ, H.C. (2011) Comparing different multivariate
calibration methods for the determination of soil organic carbon pools with visible to near
infrared spectroscopy. Geoderma, 166: 198–205.
228. Vohland, M. and Emmerling, C. (2011) Determination of total soil organic C and hot water-
extractable C from VIS-NIR soil reflectance with partial least squares regression and spectral
feature selection techniques. Eur. J. Soil Sci., 62: 598–606.
186 J. M. Soriano-Disla et al.

229. Xie, H.T., Yang, X.M., Drury, C.F., Yang, J.Y., and Zhang, X.D. (2011) Predicting soil organic
carbon and total nitrogen using mid- and near-infrared spectra for Brookston clay loam soil in
Southwestern Ontario, Canada. Can. J. Soil Sci., 91: 53–63.
230. Cambule, A.H., Rossiter, D.G., Stoorvogel, J.J., and Smaling, E.M.A. (2012) Building a near
infrared spectral library for soil organic carbon estimation in the Limpopo National Park,
Mozambique. Geoderma, 183–184: 41–48.
231. Fuentes, M., Hidalgo, C., González-Martı́n, I., Hernández-Hierro, J.M., Govaerts, B., Sayre,
K.D., and Etchevers, J. (2012) NIR spectroscopy: An alternative for soil analysis. Commun.
Soil Sci. Plant Anal., 43: 346–356.
232. Northup, B.K. and Daniel, J.A. (2012) Near infrared reflectance-based tools for predicting soil
chemical properties of Oklahoma grazinglands. Agron. J., 104: 1122–1129.
233. McDowell, M.L., Bruland, G.L., Deenik, J.L., Grunwald, S., and Knox, N.M. (2012) Soil total
carbon analysis in Hawaiian soils with visible, near-infrared and mid-infrared diffuse reflectance
spectroscopy. Geoderma, 189–190: 312–320.
Downloaded by [CSIRO Library Services] at 04:00 17 February 2014

234. Xie, X.L., Pan, X.Z., and Sun, B. (2012) Visible and near-infrared diffuse reflectance spec-
troscopy for prediction of soil properties near a copper smelter. Pedosphere, 22: 351–366.
235. Yang, H., Kuang, B., and Mouazen, A.M. (2012) Quantitative analysis of soil nitrogen and
carbon at a farm scale using visible and near infrared spectroscopy coupled with wavelength
reduction. Eur. J. Soil Sci., 63: 410–420.
236. Yang, X.M., Xie, H.T., Drury, C.F., Reynolds, W.D., Yang, J.Y., and Zhang, X.D. (2012)
Determination of organic carbon and nitrogen in particulate organic matter and particle size
fractions of Brookston clay loam soil using infrared spectroscopy. Eur. J. Soil Sci., 63: 177–188.
237. Ludwig, B., Khanna, P.K., Bauhus, J., and Hopmans, P. (2002) Near infrared spectroscopy of
forest soils to determine chemical and biological properties related to soil sustainability. For.
Ecol. Manage., 171: 121–132.
238. Morón, A. and Cozzolino, D. (2002) Application of near infrared reflectance spectroscopy for
the analysis of organic C, total N and pH in soils of Uruguay. J. Near Infrared Spectrosc., 10:
215–221.
239. Reeves, J., III, McCarty, G., and Mimmo, T. (2002) The potential of diffuse reflectance spec-
troscopy for the determination of carbon inventories in soils. Environ. Pollut., 116: S277–S284.
240. Coûteaux, M.M., Berg, B., and Rovira, P. (2003) Near infrared reflectance spectroscopy for
determination of organic matter fractions including microbial biomass in coniferous forest
soils. Soil Biol. Biochem., 35: 1587–1600.
241. Bloesch, P.M. (2012) Prediction of the CEC to clay ratio using mid-infrared spectroscopy. Soil
Res., 50: 1–6.
242. Song, Y., Li, F., Yang, Z., Ayoko, G.A., Frost, R.L., and Ji, J. (2012) Diffuse reflectance
spectroscopy for monitoring potentially toxic elements in the agricultural soils of Changjiang
River Delta, China. Appl. Clay Sci., 64: 75–83.
243. Barthès, B.G., Brunet, D., Brauman, A., Fromin, N., Lensi, R., Volant, A., Laclau, J.P., Blavet,
D., and Chapuis-Lardy, L. (2010) Determination of potential denitrification in a range of tropical
topsoils using near infrared reflectance spectroscopy (NIRS). Appl. Soil Ecol., 46: 81–89.

Potrebbero piacerti anche