Вы находитесь на странице: 1из 24

Applied Spectroscopy Reviews

ISSN: 0570-4928 (Print) 1520-569X (Online) Journal homepage: http://www.tandfonline.com/loi/laps20

Proximal and remote sensing techniques for


mapping of soil contamination with heavy metals

Tiezhu Shi, Long Guo, Yiyun Chen, Weixi Wang, Zhou Shi, Qingquan Li &
Guofeng Wu

To cite this article: Tiezhu Shi, Long Guo, Yiyun Chen, Weixi Wang, Zhou Shi, Qingquan Li &
Guofeng Wu (2018): Proximal and remote sensing techniques for mapping of soil contamination
with heavy metals, Applied Spectroscopy Reviews, DOI: 10.1080/05704928.2018.1442346

To link to this article: https://doi.org/10.1080/05704928.2018.1442346

Published online: 06 Mar 2018.

Submit your article to this journal

Article views: 52

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=laps20
APPLIED SPECTROSCOPY REVIEWS
https://doi.org/10.1080/05704928.2018.1442346

Proximal and remote sensing techniques for mapping


of soil contamination with heavy metals
Tiezhu Shi a,b,c, Long Guod, Yiyun Chene, Weixi Wanga,b, Zhou Shif, Qingquan Lia,
and Guofeng Wu a,c
a
Key Laboratory for Geo-Environmental Monitoring of Coastal Zone of the National Administration of Surveying,
Mapping and GeoInformation & Shenzhen Key Laboratory of Spatial Smart Sensing and Services, Shenzhen
University, Shenzhen, China; bSchool of Architecture & Urban Planning, Shenzhen University, Shenzhen, China;
c
College of Life Sciences and Oceanography, Shenzhen University, Shenzhen, China; dCollege of Resources and
Environment, Huazhong Agricultural University, Wuhan, China; eSchool of Resource and Environmental Sciences,
Wuhan University, Wuhan, China; fInstitute of Applied Remote Sensing and Information Technology & College of
Environmental and Resource Sciences, Zhejiang University, Hangzhou, China

ABSTRACT KEYWORDS
Heavy metal soil contamination is a severe environmental problem Environmental;
globally, and its mapping is vital for environmental managers and spectroscopy
policymakers to determine its distributions and hotspots. This paper
reviewed multiple proximal and remote sensing spectroscopy
for convenient and inexpensive method of obtaining soil reflectance
spectroscopy or environmental covariates, which can be used
for mapping heavy metal soil contamination. Furthermore, spatial
prediction using proximal remote-sensed data and environmental
covariates was discussed. We suggested that mapping of the spatial
distributions of metal species may be important due to the different
bioavailabilities and toxicities of various species. The assimilation
of multiple proximal/remote-sensed sensors may promote the
horizontal and vertical mapping of soil heavy metals. Moreover,
combining the advantages of satellite and unmanned aerial vehicle
-based hyperspectral imaging systems will facilitate the development
of a space–aeronautic incorporation hyperspectral observation
technology that can monitor soil environment rapidly and accurately
at a large scale.

1. Introduction
Soil, as an open system, exchanges matter and energy with its surrounding atmosphere, bio-
sphere, and hydrosphere; soil also provides fundamental natural resources for the survival of
most terrestrial life (1). These functions depend on the balances of soil structure, composi-
tion, and the chemical, biological, and physical properties. However, accumulated soil con-
taminants may disrupt these balances and further threaten the health of plants, animals, and

CONTACT Guofeng Wu guofeng.wu@szu.edu.cn Key Laboratory for Geo-Environmental Monitoring of Coastal Zone
of the National Administration of Surveying, Mapping and GeoInformation & Shenzhen Key Laboratory of Spatial Smart Sens-
ing and Services, Shenzhen University, No. 3688 Nanhai Rd., Shenzhen City 518060, Guangdong, China.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/laps.
© 2018 Taylor & Francis Group, LLC
2 T. SHI ET AL.

the human race. For example, the presence of heavy metals in soil can reduce the seed
sprouting and root growth of some plant species, such as peach trees, cotton, citrus, and soy-
beans, thereby causing slow growth and eventually death (2). Moreover, some crop plants
potentially uptake and accumulate hazardous contaminants in their edible parts, and they
may poison humans through food chains (3).
Approximately 1.4 £ 107 sites have been contaminated with heavy metals or organics
worldwide because of anthropogenic activities, such as mining, urbanization, and agricul-
tural and industrial processes (4). In China, approximately 8.3% of 120 million hectares of
arable land are contaminated (5). In addition, approximately 65% of all Chinese cities have
high or extremely high levels of heavy metal contamination in soils and road dusts (6). Given
the threats of soil contamination to food security and human health, the identification and
remediation of contaminated sites are increasingly needed (5).
Atomic absorption spectroscopy (AAS) and inductively coupled plasma mass spectrome-
try (ICP-MS) are 2 main spectroscopic techniques used to determine the metal concentra-
tions of soil samples, and these measurements can be considered closest to their true values
(7, 8). In contrast to limited soil samples with AAS and ICP-MS methods, spatial maps are
highly useful for environmental managers and policymakers to determine the distribution
and hotspots of heavy metal soil contamination. Fortunately, according to the measured val-
ues from AAS and ICP-MS and geostatistical approaches, mapping the spatial distribution
of heavy metals within polluted sites is commonly feasible (9). Nevertheless, some processes
in AAS and ICP-MS, such as grinding and acid digestion of soil sample into a solution, are
expensive, environmentally toxic, and time consuming, especially when many measured val-
ues are required for improving spatial interpolation accuracy. Although Jenny’s model, also
known as CLORPT model (10), requires less reference soil samples for spatial predictions
(11), lacking data source for deriving environmental covariates restricts its application in
digital soil mapping.
The developing technologies of proximal and remote sensing have become efficient meth-
ods for gaining large amount of geographical data (12). The development of new tools in the
statistical fields and later thrived machine learning are providing new ways to analyze these
data (13). All these technologies have considerably changed soil mapping nowadays (14)
and offered new methodologies for digital mapping of soils with heavy metal contamination.
Given the importance of mapping soil contamination, reviewing the progress in digital
soil mapping of heavy metals and further promoting its studies and applications are essen-
tial. Therefore, this paper provides an introduction about multiple spectroscopic sensors for
convenient and inexpensive acquiring of soil information or environmental covariates and
further discusses the challenges and potential research directions in mapping soils with
heavy metal contamination.

2. Spectroscopic technologies
Multiple spectroscopic technologies, such as proximal and remote sensing, can provide inex-
pensive and accessible data sources for the spatial prediction of soil heavy metals. The appli-
cation scenarios of these techniques in mapping soil heavy metals can be classified into
3 categories:
i) Proximal sensing technologies, including X-ray fluorescence spectrometry and visible
and near-infrared reflectance spectroscopy, provide the spectra of soil samples. Soil
APPLIED SPECTROSCOPY REVIEWS 3

heavy metals can be determined from these spectra using a prediction model; after-
ward, spatial maps are produced by directly interpolating the retrieved values (15).
ii) Hyperspectral remote sensing technologies, such as airborne or satellite-borne hyper-
spectral sensors, acquire the hyperspectral images of land soils. Hyperspectral spec-
troscopy of soils can be derived from the image pixels. Unlike proximal sensing,
which acquires the reflectance spectroscopy of certain points, hyperspectral imaging
sensors obtain the planar spectral information of soils on a large scale.
iii) Other remote sensing technologies, such as multispectral images and light detection
and ranging (LiDAR), can be used as data sources to extract environmental covariates,
including climate, vegetation, land uses, and topography factors. The covariates can be
integrated with CLORPT model as parameters for spatial prediction of soil heavy
metals.

2.1. Proximal sensing


2.1.1. X-ray fluorescence spectrometry
When atom is irradiated or excited by X-rays, an inner shell vacancy is created, with an
electron hole left in the inner shell. Almost instantaneously, an electron from a higher
energy shell will drop down to fill the vacant inner shell as the atom relaxes to ground
state. This process releases photons with energy in the X-ray region of electromagnetic
spectrum equivalent to the energy difference between the 2 shells (Figure 1) (16). The
detection of specific fluorescent photons enables the qualitative and quantitative analyses
of most elements in samples (17). Radioisotope sources and X-ray tubes are frequently
used to irradiate a sample, and the most commonly used radioisotope sources include Fe-
55, Co-57, Cd-109, and Am-241. Each of these radioisotope sources releases radiation at
specific energy levels. Thus, 2 or 3 sources are adopted to excite the entire range of ele-
ments of interest in soil analysis. Additionally, radioisotopes are often inconvenient, expen-
sive, and with hazardous excitation source for portable systems because of their half-life
characteristics (18). Hence, low dissipative X-ray tube excitation sources have become pop-
ular in recent years. X-ray fluorescence spectrometers can be divided into 2 categories:
wavelength and energy dispersive X-ray fluorescence. Wavelength dispersive spectrometers
disperse emitted photons by using crystal, diffraction, or special (geometric) dispersion. In
energy dispersion, the X-ray line separation, which is based on photon energies, is accom-
plished by electronic dispersion with a pulse height analyzer (16). Currently, energy

Figure 1. Mechanism for X-ray fluorescence of an atom (16). Reproduced with permission of Publisher.
4 T. SHI ET AL.

dispersive X-ray fluorescence is an extensively used portable spectrometer because of its


simple instrumentation and improved performance (16).
Empirical and fundamental parameter calibrations are 2 major approaches for calibrating
X-ray fluorescence applications in soil samples. Empirical calibrations are typically based on
a set of previously collected site-specific calibration references, which have been certified by
reliable independent laboratory methods such as ICP-MS and AAS (16). A large number of
representative references may be vital for improving the accuracy of the empirical calibration
model. In the absence of site-specific calibration requirement, the fundamental parameter
technique utilizes a physical theory to predetermine inter-element coefficients, converts
background, and overlaps corrected net intensities into concentrations. Adjustments based
on certified reference materials may be necessary to correct errors and biases caused by
uncertainties in data used in the fundamental parameter model (16, 18). The fundamental
parameter technique requires only one certified reference material to calibrate X-ray fluores-
cence system compared with that of empirical approach; all relevant elements are included
in the concentration calculation, and the calibrated model is not site-specific (16).
X-ray fluorescence methodology has been accepted by the environmental community as a
viable and cost- and time-effective approach for analyzing hazardous materials (16). Nowa-
days, X-ray fluorescence is a widely accepted approach for quantitative or qualitative analy-
ses of heavy metals in soil and sediments (16). However, the analysis accuracy of X-ray
fluorescence may be adversely affected with several properties of soil samples, such as mois-
ture, organic content, grain-size distribution, and surface characteristics (19). Sample prepa-
ration procedures, including drying, grinding, sieving, homogenizing, and/or compacting
sample materials, may reduce these effects and improve analysis accuracy.

2.1.2. Visible and near-infrared reflectance spectroscopy


When the radiation containing all relevant frequencies in the visible and near-infrared range
is directed to the soil sample, the individual molecular bonds of soil constituents will vibrate,
either by bending or stretching. The individual molecular bonds will absorb light at various
degrees, with a specific energy quantum corresponding to the difference between 2 energy
levels (Figure 2a) (20). The weak overtones and combinations of these fundamental vibra-
tions due to the stretching and bending of NH, OH, and CH groups occur in the near-infra-
red (700–2500 nm) and electronic transitions in the visible range (400–700 nm) (Figure 2b)
(21). The energy quantum is directly related to frequency and inversely related to wave-
length; thus, the absorption features at specific wavelengths can be used for quantitative
analysis of the soil constituent concentrations (22).
Given the low concentrations and overlapping absorptions of soil constituents, visible and
near-infrared spectra are largely nonspecific, relatively weak, and broad. Therefore, the anal-
ysis of visible and near-infrared spectra requires the use of multivariate chemometric techni-
ques to extract useful information for soil property estimations mathematically.
Preprocessing techniques, including Savitzky–Golay smoothing, derivative transformation,
standard normal variate, and multiplicative scatter correction, are commonly used to reduce
random noise, baseline drift, and multiple scattering effects in spectra (23). Feature selection
techniques, such as successive projection algorithm, uninformative variable elimination, and
genetic algorithm, are often applied to remove uninformative spectral bands and select opti-
mal spectral variable subsets for establishing regression models (24, 25). Various data mining
techniques, such as principal component regression, partial least squares regression, artificial
APPLIED SPECTROSCOPY REVIEWS 5

Figure 2. (a) Molecular level transition, (b) soil visible and near-infrared (400–2400 nm) spectra showing
approximately where the combination, first, second, and third overtone (OT) vibrations occur (21, 20).
Reproduced with permission of Publisher.

neural network, multivariate adaptive regression splines, and support vector machine, are
used to train models from spectral data for estimating soil properties. With the progress in
chemometric methods, visible and near-infrared reflectance spectroscopy has become a
mature technology for predicting numerous soil physical, chemical, and biological proper-
ties, including metals (21, 26).
Currently, researchers dedicate their efforts to establish national or global soil spectral
libraries and develop field systems for soil spectra measurement. Viscarra Rossel et al. (27)
developed the largest and most diverse global soil spectral library; they indicated the useful-
ness of global spectra for predicting soil properties, including soil organic and inorganic car-
bon, clay, slit, sand, iron contents, cation exchange capacity, and pH. They suggested that
new contributions to the library will help develop a dynamic and easily updatable database
with improved global coverage. They also hoped that the use of this database will deepen
our understanding about soil for sustainable management and extend the research outcomes
of soil, earth, and environmental sciences. The spectral library is useful for the prediction of
soil properties in regions where the measurements of these qualities are lacking or consider-
ably expensive to perform using conventional laboratory methods (21, 27). Nevertheless,
spectral libraries should contain sufficient samples to adequately describe the soil variability
6 T. SHI ET AL.

of the site where the models will be used (28). To improve the local predication by using
spectral libraries, 2 modeling strategies have been developed: spiking modeling by adding
typical local samples into the spectral library (29, 30) and local statistical modeling, such as
locally weighted regression (31, 32), LOCAL algorithm (33), and spatially constrained local
partial least squares regression (34).
Recently, field measurement systems, including on-the-go and in situ systems, have been
attracting increasing attention. Stenberg et al. (20) summarized that on-the-go systems
include a plough equipped with a fiber optic probe and an artificial light protected with a
sapphire glass at the bottom. The probe penetrates into the soil and rhythmically acquires
soil spectra when the plough is powered with a mechanical transmission. In situ systems
obtain soil spectra in the field with a contact probe. Consequently, the tedium of drying,
grinding, and sieving soil samples will be avoided. For example, Li et al. (35) used a handheld
and high-intensity contact probe to measure the spectra of soil profiles in situ. Viscarra Ros-
sel et al. (36) developed an automated soil core sensing system to acquire the spectra of soil
profiles. These in situ systems improve the efficiency of soil spectral measurement and
address the problem on acquiring soil measurements at depth.
Visible and near-infrared reflectance spectroscopy can serve as a rapid, noninvasive, cost
effective, and environmentally friendly alternative for analyzing soil contamination by heavy
metals (23). Heavy metals, including As, Cd, Co, Cr, Cu, Hg, Pb, and Zn within suburban,
mining region, floodplain, and agricultural areas, have been successfully predicted by using
reflectance spectra (37–43). Spectral prediction is based on surrogate correlations between
the concentrations of heavy metals and those of O-, C-, and N-bearing compounds (44, 45);
these statistical correlations may be site-specific, and they may depend on soil-forming cir-
cumstances. Given the surrogate relations, Baveye and Laba (46) suggested that the applica-
tion of this technique will be limited for monitoring of soil contamination by heavy metals.
Additionally, other factors, such as soil moisture content, surface roughness, and shallow
penetration depth less than a millimeter, will affect the applications of visible and near-infra-
red reflectance spectroscopy (46). However, we consider that the effects of soil moisture con-
tent and surface roughness can be eliminated by using proper preprocessing methods, such
as orthogonal signal correction (47, 48); the in situ and on-the-go measurements of soil pro-
files may also resolve the shallow penetration depth problem. With consideration of the
capacity of visible and near-infrared reflectance spectra in accurately predicting soil organic
matter, pH, clay, silt, sand, iron, and manganese oxide, which are related to heavy metals, we
believe that reflectance spectra may serve as covariates for promotion of the spatial predic-
tion of soil heavy metals.

2.2. Remote sensing


2.2.1. Hyperspectral remote sensing
Hyperspectral image (Figure 3a) is acquired by a hyperspectral imaging sensor on airborne
or spaceborne platform. After geometric, radiometric, and atmospheric corrections, the visi-
ble and near-infrared reflectance spectra of image pixels are obtained (Figure 3b). Prepro-
cessing and modeling methods are used to calibrate an empirical model and consequently
relate these reflectance spectra to the soil property contents of training samples. Thus, the
soil property content in each pixel is estimated by using calibrated empirical model and
reflectance spectroscopy.
APPLIED SPECTROSCOPY REVIEWS 7

Figure 3. Airborne hyperspectral cube acquired by Headwall Micro-Hyperspec sensors (a) and the reflec-
tance spectra derived from image pixels (b).

Airborne or spaceborne hyperspectral images have been applied to map the spatial distri-
bution of heavy metals. For example, Kemper and Sommer (49) used airborne hyperspectral
images obtained by a HyMap sensor (http://www.intspec.com/) to map Pb and As contami-
nations in the Guadiamar floodplain, Andalusia; Choe et al. (37) used spectral parameters,
such as the ratio of 1344 nm to 778 nm, absorption area at 2200 nm, and absorption depth
at 500 nm derived from HyMap data, to map the spatial distribution of Pb, Zn, and As in
the Rodalquilar mining area, SE Spain. Furthermore, Wu et al. (50) indicated that the simu-
lated HyMap provides satisfactory results in mapping heavy metals in Nanjing City, China.
These direct predictions using hyperspectral images require signals from bare soils, which
are often disturbed with vegetation coverage. Airborne hyperspectral images can be acquired
in winter or early spring or in crop rotation within agricultural areas when vegetation cover-
age is low (51, 52). In addition, Diek et al. (53) created multitemporal composites of airborne
images to increase the spatial coverage of bare soils and further support the digital soil map-
ping of multivariate soil properties, including heavy metals.

2.2.2. Other remote sensing data sources for extracting environmental covariates
Multispectral remote sensing can be used to extract environmental covariates for predicting
soil heavy metals. National Oceanic and Atmospheric Administration (https://www.climate.
gov/maps-data) provides the maps of climate indicators derived from meteorological satellite
data and ground monitoring station data on global scale, such as average monthly tempera-
ture and precipitation. Multispectral remote-sensed images acquired from MODIS, Landsat,
SPOT, and Gaofen satellites can be used to derive organism indices, such as vegetation indi-
ces and land use patterns. Wilford et al. (54) utilized MODIS time-series data from 2000 to
8 T. SHI ET AL.

2008 to obtain 12 coefficients based on enhanced vegetation index, which reflect vegetation
changes in greenness; these coefficients are also used as environmental covariates to predict
soil heavy metals. Lado et al. (55) adopted enhanced vegetation index derived from MODIS
1 km images as environmental data for spatial prediction of heavy metals in the European
Union. Peng et al. (56) used multivariate spectral indices derived from Landsat 8 images
(enhanced vegetation index, normalized difference vegetation index, and soil-adjusted vege-
tation index) and other environmental variables to model and map the spatial distribution
of As, Cr, Ni, Cu, Pb, and Zn in Qatari soils. ASTER adopting stereo-imaging technology
provides global digital elevation model with a spatial resolution of 30 m. Moreover, Shuttle
Radar Topography Mission is claimed to provide digital elevation models with a vertical
absolute accuracy of less than 16 m and an absolute horizontal accuracy of 20 m (57). In
addition, fine-resolution digital elevation models can be derived from LiDAR technology,
where the LiDAR points are interpolated using triangulation method (58). According to dig-
ital elevation models, various terrain indices, including elevation, aspect, slope, compound
topographic index, stream power index, and wetness index, can be obtained by using digital
terrain analysis technology. With the used of LiDAR-derived digital elevation models, Peng
et al. (56) calculated various geomorphology data (elevation, slope gradient, slope aspect,
and distance to drainage line) for spatial prediction of heavy metal distribution.

3. Spatial prediction methods


X-ray fluorescence spectrometry and visible and near-infrared reflectance spectroscopy can
provide information about heavy metal contents in sample sites. Remote sensing techniques
offer data sources for extracting environmental covariates. Nonetheless, the use of informa-
tion in sample sites or the environmental covariates derived from proximal or remote sens-
ing to produce the maps of heavy metal distribution remains a challenge. In this section,
theories and statistical methods for spatial prediction, including CLORPT model, spatial
regression, and spatial interpolation, are presented.

3.1. Theoretical bases for spatial prediction


Current quantitative predictive models for soil mapping are generally based on 2 theories. One
theory is Jenny’s or CLORPT model, and it regards soil as the product of the joint action of cli-
mate (cl), organisms (o), relief (r), parent material (p), and time (t) (10), which can be described
as S D f(cl, o, r, p, t,…). Implicitly, climate includes rainfall, temperature, and solar radiation;
organisms refer to vegetation cover and types, land use, and anthropological activities; relief
means topography, such as slope aspect, elevation, and slop angle; parent material, such as rock
type, is the original supply of soil mineral elements; and time is often a theoretical or hypothetical
span for soil development. The choice of soil-forming factors may depend on the soil surveyors’
prior knowledge. If close relations exist between environment and soil, predicting soil from easily
obtainable environmental covariates is preferred over obtaining soil itself. CLORPT model has
been used by innumerable surveyors worldwide as a qualitative criterion for exploring the envi-
ronmental factors vital to soil pattern formation (14).
Environmental scientists know intuitively that the values of environmental properties are
naturally related, that is, autocorrelated at some spatial scales. Geostatistics expresses this
intuitive knowledge quantitatively and subsequently applies it for prediction. Generally,
APPLIED SPECTROSCOPY REVIEWS 9

researchers can consider the soil at certain location (x, y) by depending on geographic coor-
dinates (x, y) and the soil at neighboring locations, namely, (xCu, yCv), S(x, y) D f ((x, y), S
(xCu, yCv)), where u and/or v determine the dependence of this function (14). Inevitable
errors occur in prediction, but they can be minimized and estimated by quantifying the spa-
tial autocorrelation at the scale of interest. When available soil-forming factors fail to obtain
a satisfactory soil spatial prediction, this neighborhood law may arise as an alternative for
soil mapping (14).

3.2. Related mathematical and geostatistical methods


Jenny (10) attempted to express the relationship between soil and its forming environment
mathematically; nonetheless, this attempt is largely unprofitable because he did not know
how to relate multiple soil variables with environmental covariates at a time. The ice presents
no melting until the development of computer and canonical correlation. Canonical correla-
tion calculates a canonical variable, which is a linear function of variables, and subsequently
relates the 2 sets’ canonical variables to present their relations (59). Webster (60) first
adopted computer-aided canonical correlation analysis to relate soil and environmental data
for soil surveys in Britain and Australia. The universal application scene of Jenny’s model
uses ancillary environmental data to predict the classes and the physical and chemical prop-
erties of soils. Univariate or multivariate linear regressions were adopted by most of the pre-
vious studies (61–63). For instance, Moore et al. (62) used linear models to predict soil
attributes (organic matter content, extractable phosphorous, pH, and texture) from terrain
variables derived from a 15-m digital elevation model in Northeastern Colorado. However,
the linear model is a 2-edged sword (64). On the one hand, linear model is simple, and it
can easily interpret the relationships between variables. On the other hand, nonlinear model-
ing problems, such as non-normality and constant variance of variables, often conflict with
the underlying assumptions of linear model. Given their advantages in handling nonlinear
problems, generalized linear model (64), generalized additive model (65), support vector
machine (66), artificial neural network (67), and regression forests (65, 68) are used to
accommodate the nonlinear relations between soil and environment. Machine learning tech-
niques often result in more robust regression models than those of linear regressions; how-
ever, they are complex and difficult to interpret.
Geostatistical method, a subset of traditional statistic functions, is based on the theory of
regionalized variable (69, 70), and it primarily deals with spatial data and accounts for spatial
autocorrelation using spatial interpolator (71). Since the first introduction of ordinary point
or block kriging to soil studies by Burgess and Webster (72), it has been used in soil mapping
by spatially interpolating soil property values at unvisited sites from field-collected data,
including soil classification, pollution, reclamation, trace element deficiencies, salinity, and
fertility (71, 73). Moreover, other interpolation methods, such as inverse distance weighting,
radial basis function, and back-propagation neural network, are adopted for geostatistical
prediction of the spatial distributions of soil properties (74–76). One limitation of univariate
geostatistical technique is the assumption of spatial autocorrelation, which is excessively
data dependent, and it often requires a large amount of samples, especially for areas with
high spatial variability (77). Importantly, this assumption may be poor in complex terrain
where soil-forming processes are also complex (73). With the knowledge of soil-forming
process, considering available and cheap ancillaries in modeling is economic and logistic,
10 T. SHI ET AL.

but geostatistics, such as kriging, fail (73, 78). Therefore, an appropriate combination of uni-
or multivariate analysis using environmental factors and geostatistics, that is, a hybrid tech-
nique, is proposed to solve these limitations. This technique supposes that a vector of soil
property Z, determined at locations in a region X D x1,…, xN, consists of 3 components:

Z ðxÞ D m C Z1 ðxÞ C eðxÞ; (1)

where Z1(x) is the spatially dependent component, e is the spatially independent residual
error term, and m exhibits a trend or changing drift that may depend on CLORPT factors
(73). To accommodate the trend m, ordinary univariate kriging is inappropriate, and several
hybrid techniques have been designed. Cokriging is a multivariate extension of kriging that
includes many available and cheap CLORPT variables, such as climate, topography, and
time, in the prediction process (73). Cokriging may improve prediction when CLORPT vari-
ables are unavailable at all grid nodes, and their numbers are low (79). Universal kriging, a
combination of multiple linear regression (CLORPT model) and ordinary kriging using geo-
graphical coordinates, has been widely used to accommodate the trend in a soil variable
(80). Kriging with external drift is similar to universal kriging, and it represents the trend by
using ancillary drift variables, such as digitized covariates derived from digital elevation
model, rainfall data, or scanned images (81). Kriging with external drift has become an
uncommonly used method in soil science; nevertheless, with the increased availability of
remotely sensed data, it may be well used, along with regression–kriging (82). Regression–
kriging combines the trend predicted by uni- or multivariate regression models with the
regression residuals spatially interpolated from a kriging model (83). This method assumes
that the target soil variable trend (m in Eq. 1) is accounted for by regression model, and
model residuals represent spatially dependent component (Z1(x) in Eq. 1) (73).
Hybrid techniques, such as universal kriging, kriging with external drift, and regression–
kriging, have been widely used in spatial prediction of various soil properties, including organic
matter, pH (84), depth of solum and bedrock, gravel, clay (85), and heavy metals (86), and a
summary was provided by McBratney et al (14). However, some confusions exist because of the
same name for different methods and different names used by different researchers for the
same approach. Hengl et al. (79) attributed these confusions to the different goals of various
professions; for example, geostatisticians consider these methods special cases of interpolation
technique, and statisticians consider kriging a case of regression analysis with spatially corre-
lated data. Furthermore, Hengl et al. (79) distinguished the application scenarios of these meth-
ods: universal kriging is applied to cases where only the trend is modeled as a function of
coordinates; kriging with external drift is used when the drift or trend is defined externally by
using some ancillary variables, instead of using coordinates; regression–kriging can fit the trend
and residuals separately and subsequently add them, and its advantage is that it can be easily
combined with generalized additive models and regression trees.
A recent approach for spatial modeling of environmental covariates, that is, geographi-
cally weighted regression (GWR), has attracted the attention of environmental experts (87).
GWR extends the framework of traditional regression, considers the spatial location of data
points, and estimates the local regression coefficients of environmental covariates at different
locations rather than global coefficients (88). The regression coefficients of GWR can reflect
the spatial varying relationships between soil properties and environmental covariates. Addi-
tionally, GWR kriging (GWRK) (89), the combination of GWR and kriging, considers the
APPLIED SPECTROSCOPY REVIEWS 11

regression residuals of GWR model and improves the spatial prediction of soil properties at
regional scale (90–92). In GWRK modeling, the target variable trend is modeled with GWR
and environmental covariates, and the residuals are interpolated with kriging and added to
the trend because they help explain the variation in target variables across space. GWRK is
an efficient method of residual estimation for trend prediction (90).

4. Perspectives
4.1. Heavy metal species
To date, many studies have focused on spatial prediction of the total heavy metal concentra-
tions and evaluation of the bioavailability and toxicity of soil contaminated with heavy met-
als (Table 1). However, the bioavailability and toxicity of metals are largely determined by
their species, and different species result in a variety of effects on environment, thereby
affecting the migration and circulation of metals in nature (93). Metal species indicate the
existing forms of metal in soils as ion or molecule, including valence, compound, bound,
and structural states. No set definition or classification exists for metal species. Tessier et al.
(94) divided soil metals into 5 species, including exchangeable, bound to carbonates, bound
to iron and manganese oxides, bound to organic matter, and residual, which are the most
widely adopted classifications.

Exchangeable
Metals that are absorbed by clays, hydrated oxides of iron and manganese, and humic acids
are sensitive to environmental changes and easy to migrate and transfer; plants can also
uptake these metals (93).

Bound to carbonates
Metals associated with soil carbonates can be susceptible to changes in pH and released due
to the decrease in pH (95).

Bound to iron and manganese oxides


Metals absorbed by or coprecipitated with surfaces of iron and manganese oxides are highly
affected with soil pH and redox conditions (93, 94).

Bound to organic matter


Metals bound to various forms of organic matter, such as living organisms, detritus, and
coatings on mineral particles, can be released into soluble metals when organic matter is
degraded under oxidizing conditions (94).

Residual
Metals existing in the crystal structure of primary and secondary minerals may not be released in
solution over a reasonable time span under the conditions normally encountered in nature (94).
Given the different bioavailabilities and toxicities of various species, mapping the spatial
distributions of metal species may be more important than that of total concentrations.
According to soil sampling and subsequent Tessier sequential extraction procedure with
AAS, mapping metal species can be performed by using spatial interpolation methods. X-ray
12

Table 1. Literatures for spatial prediction of heavy metals in soils.


Ancillary data for Study area
Reference Metals Predictive methods modeling Sampling methods No. of samples Locations (km2)

Guo et al. (100) As, Pb, Zn, Cu Kriging — Random sampling 63 Urban areas, Yibin, 68
Southwest
T. SHI ET AL.

China
Maas et al. (101) Cd, Cr, Cu, Pb, Zn Cokriging Soil type, pH, Mn, Fe, etc. Grid sampling 101 Annaba, a 101
Mediterranean
city of Algeria
Navas and Machin (102) Cr, Cu, Ba, As, Sb, Hg, Sn, Mn, Ordinary block kriging — Random sampling 133 Aragon, northeast 45000
Fe, Al, Zn, Ni, Co, Cd and Spain
Pb
Imperato et al. (103) Cu, Cr, Pb, Zn Kriging — Grid sampling 173 Naples city urban 43
area, Italy
Xie et al. (74) Cu, Pb, Cd Inverse distance — Random sampling 137 Tongzhou district, 605
weighting, ordinary Beijing, China
kriging, radial basis
functions
Zhang et al. (104) Cd Ordinary kriging — Random sampling 260 Guangdong 1.78 £ 105
Province, China
Ordonez et al. (105) Cd, Zn, Hg Kriging — Grid sampling, and 16 112 Aviles, Northern 7
randomed samples Spain
in each grid
Martinez-Garcia et al. Pb, Cd, Cu, Zn, Al Spatial interpolation — Stratified sampling 112 Cartagena area in —
(106) southeastern
Spain
McGrath et al. (107) Pb, Zn, Cu, Cd, As Ordinary kriging — Grid sampling 223 Silvermines area, 32.64
Ireland
Saby et al. (108) Pb Ordinary kriging, — Grid sampling 67 Paris, France 25000
lognormal kriging
Romic and Romic (109) Cd, Cu, Fe, Mn, Ni, Pb, Zn Kriging — Grid sampling 331 Zagreb area, 331
Croatia
Martin et al. (110) Cd, Cr, Cu, Hg, Ni, Pb and Zn Ordinary kriging — Grid sampling 624 Ebro basin, Spain 40000
Lee et al. (111) Cd, Co, Cr, Cu, Ni, Pb, Zn Inverse distanced — 5 samples/km2 in urban 450 Hongkong Island 80.3
weighting areas and 2 samples/ area
km2 in the suburban
and country park
sites
Sun et al. (112) Cr, Ni, Cu, Zn, and Pb Cokriging Soil organic matter, pH, Random sampling 114 Dehui County, 1350
Clay, Silt, Sand, heavy Northeast China
metals
Facchinelli et al. (113) Cr, Co, Ni, Cu, Zn, Pb Kriging — Random sampling 98 Piemonte, 25400
northwest Italy
Li et al. (114) Cd, Co, Cr, Cu, Ni, Pb, Zn Kriging — Grid sampling, and 5 152 Kowloon Peninsula, 46.9
random samples in Hongkong
each grid
Bou Kheir et al. (115) Ni, Cr, Cd, As Regression tree Parent material, slope Random sampling in 200 Northern part of 195
gradient, proximity to different land uses Lebanon
roads, etc.
Chen et al. (116) Cd Kriging — Random sampling 76 Fenghui, Xi’an City, 14.27
Shaanxi
Province, China
Chen et al. (86) Cr, Cu, Hg, Ni, Pb, and Zn Cokriging Soil spectra, soil Random sampling 52 Fenghui, Xi’an City, 14.27
properties, such as pH, Shaanxi
organic matters Province, China
Wilford et al. (54) Cr Decision tree-based Environmental factors: Grid sampling 1026 Sir Samuel, 16000
modeling terrain, climate, Western
geology, vegetation Austrialia
Zhang (117) Cu, Pb, Zn, As Inverse distance — Grid sampling 166 Galway, Ireland 54
weighting
Lado et al. (55) As, Cd, Cr, Cu, Ni, Pb, Zn, Hg Regression–kriging topographic indexes, land Grid sampling 1588 European Union
cover, geology,
vegetation indexes,
night lights images
and earth quake
magnitudes
Peng et al. (56) As, Cr, Ni, Cu, Pb and Zn Cubist algorithm Spectral indices, Stratified random 300 The State of Qatar
geopedological sampling
information,
geomorphology data,
anthropogenic
information
APPLIED SPECTROSCOPY REVIEWS
13
14 T. SHI ET AL.

fluorescence spectrometry and visible and near-infrared reflectance spectroscopy may be


alternative techniques for Tessier sequential extraction procedure; however, this hypothesis
needs further exploration. Zhou and Huang (96) reported that soil properties (pH, redox,
organic matter, and colloid) and some environmental covariates (climate, hydrology, and
organisms) are important factors affecting the species of metal. As mentioned in the previous
section, soil properties can be estimated by X-ray fluorescence spectrometry and visible and
near-infrared reflectance spectroscopy. Climate, hydrology, and organism covariates can be
obtained from remote-sensed data. Therefore, the spatial distribution of metal species may be
mapped by adopting multisource proximal/remote sensed data and CLORPT model.

4.2. Multiple sensor fusion


The overtones and combination bands in near-infrared region are weak and overlapping,
thereby making spectra difficult to characterize (21, 28). In contrast to that in the near-infra-
red region, the spectral features in the mid-infrared region are frequently better resolved and
considerably more intense; more spectral features are also available in this spectral region
(26). Mid-infrared reflectance spectroscopy generally produces highly accurate prediction of
soil properties (21, 26). In addition, Siebielec et al. (40) found that mid-infrared reflectance
spectroscopy markedly outperforms near-infrared reflectance spectroscopy in predicting soil
metals, including Fe, Cd, Cu, Ni, and Zn. However, this technology is more complex and
expensive than using visible and near-infrared reflectance spectroscopy. With technical devel-
opment, the cost of mid-infrared reflectance spectroscopy decreases, which increases the pos-
sibility of in situ and on-the-go prediction of soil properties, including heavy metals (21).
Although Soriano-Disla et al. (26) proposed that the combinations of visible, near- and
mid-infrared reflectance spectroscopy generally result in worse or no prediction improve-
ment compared with those of single spectral ranges; we speculate that the combinations of
X-ray fluorescence spectrometry and visible, near-, and mid-infrared reflectance spectros-
copy may be advantageous in predicting metal species. X-ray fluorescence spectrometry can
accurately predict the total concentration of heavy metal due to its prediction mechanism at
atomic level, but it may be limited in differentiating metal species. Near- or mid-infrared
reflectance spectroscopy, which detects soil properties based on overtones and combinations
of molecular groups, can predict soil pH, organic matter, iron, and manganese, which are
critical factors controlling the forms of heavy metals in soils. The combinations of X-ray
fluorescence spectrometry and visible, near-, or mid-infrared reflectance spectroscopy will
maximize the capabilities of both technologies to predict metal species in soils. Such assump-
tion may be an interesting and notable issue that should be explored further.
The assimilation of multisource proximal/remote-sensed data may promote the horizontal
and vertical mapping of soil contaminated with heavy metals. Furthermore, the assimilation
of these data with different spatial scales will be a considerable topic in spatial prediction of
soil heavy metals. Moreover, the spatial scale will determine the spatial autocorrelation char-
acteristics of predicted variables and further ascertain the appropriate modeling approaches.

4.3. Airborne or spaceborne hyperspectral images


Unmanned aerial vehicle (UAV)-based hyperspectral imaging system is commonly used in
environmental monitoring of land surface. This system generally comprises 4 modules,
APPLIED SPECTROSCOPY REVIEWS 15

including global position system, UAV platform, pan-tilt device system, hyperspectral imag-
ing sensor, and ground control system (Figure 4). Global position system provides 3D posi-
tion information for autotrailing planned route of UAV and the geometric joint of images.
UAV platforms include fixed and rotary wings. Fixed-wing UAVs commonly present long
flight durations and fast speed, and rotary-wing UAVs (Figure 4a) offer high maneuverabil-
ity. An inertial measurement unit in UAV records the real-time data of flying attitude, which
is vital for the geometric correction of hyperspectral images. Pan-tilt-device system alleviates
the trembling of UAV in flight, which may disturb the observation stability of hyperspectral
imaging sensors and cause the geometric distortion of images. The ground control system
realizes the transmission and storage of image data and the flight control of UAV. UAV
hyperspectral imaging system provides cost-effective remote sensing hyperspectral data,
with considerably high spatial resolutions at flexible acquisition periods. UAV platform-
based remote sensing is more flexible and controllable in lying height, viewing angles, and
flight attitude than that of traditional satellite remote sensing. Therefore, this technique has
been widely applied in forestry and precision agriculture (97–99).
Several satellite hyperspectral imaging sensors will be launched in the future (Table 2).
Satellite hyperspectral imaging system can acquire images of land surface covering larger
area than that of UAV hyperspectral imaging system. Nonetheless, the radiative transfer and
atmosphere correction of satellite-borne hyperspectral images will be a difficult and rigorous

Figure 4. (a) GaiaSky-min unmanned aerial vehicle (UAV) based hyperspectral imaging system developed
by Sichuan Dualix Spectral Imaging Technology Co. Ltd, Sichuan, China, (b) the observation geometry and
ground control of UAV-hypersepctral imaging system.
16 T. SHI ET AL.

Table 2. The hyperspectral imaging satellites planned to be launched.


Hyperspectral imaging satellites Spectral resolution Spatial resolution

The fifth satellite (GF-5) of China High-resolution Earth Observation System (CHEOS) 330 spectral bands 30 m
The ALOS-3 of Japanese Hyperspectral Imager Suite (HISUI) 220 spectral bands 30 m
Italian PRecursore IperSpettrale della Missione Applicativa (PRISMA) 237 spectral bands 30 m
German Environmental Mapping and Analysis Program (EnMap) 244 spectral bands 30 m
U.S. NASA Hyperspectral Infrared Imager (HyspIRI) 210 spectral bands 60 m
Canada Hyperspectral Environment and Resource Observer (HERO) 210 spectral bands 30 m

task. Additionally, the long revisiting time of satellite sensors and the possibility of cloud
cover may restrict its observation of contamination accident at a high time resolution.
Hence, we suggest the combination of the advantages of satellite and UAV hyperspectral
imaging systems to develop a space–aeronautic incorporation hyperspectral observation
technology for rapid and accurate monitoring of soil environment at a large scale.

5. Conclusions
This article reviewed the roles of multiple proximal and remote sensing techniques in
mapping soil contaminated with heavy metals. We concluded that proximal sensing
spectroscopic sensors, such as X-ray fluorescence and visible and near-infrared reflec-
tance spectroscopy, can acquire the reflectance spectroscopy of soils. This spectroscopy
can be used to retrieve soil heavy metal contents by using a prediction model. Hyper-
spectral remote sensing can provide the planar hyperspectral information of land soils.
Other remote sensing technologies, such as multispectral images and LiDAR, can also be
used as data sources for extraction of environmental covariates, including climate, vege-
tation, land use, and topography factors. Furthermore, this paper introduced the theories
and statistical methods of spatial interpolation and CLORPT model, which support the
digital mapping of soil heavy metals by using the proximal sensing-based spectral
retrieves and remote sensing-based environmental covariates. On the basis of the prog-
ress in digital mapping of heavy metal soil contamination, we suggest the mapping of
the spatial distributions of metal species due to their different bioavailabilities and toxic-
ities. The spectroscopic sensor combination of X-ray fluorescence spectrometry and visi-
ble, near-, and mid-infrared reflectance spectroscopy may maximize the capabilities of
both technologies to predict metal species in soils. Combining the advantages of satellite
and UAV hyperspectral imaging systems will also allow the development of a space–
aeronautic incorporation hyperspectral observation technology for a rapid and accurate
monitoring of soil environment at a large scale.

Acknowledgments
This study was supported by National Natural Science Foundation of China (No. 41701476), the
National Key R&D Program of China (No. 2017YFC0506200), the Basic Research Program of
Shenzhen Science and Technology Innovation Committee (No. JCYJ20170302144323219 and
No. JCYJ20151117105543692), the Scientific Research Foundation for Newly High-End Talents
of Shenzhen University, the Shenzhen Future Industry Development Funding Program (No.
201507211219247860).
APPLIED SPECTROSCOPY REVIEWS 17

Funding
National Natural Science Foundation of China, National Key R&D Program of China, the Basic
Research Program of Shenzhen Science and Technology Innovation Committee, the Scientific
Research Foundation for Newly High-End Talents of Shenzhen University, Shenzhen Future Industry
Development Funding Program.

ORCID
Tiezhu Shi http://orcid.org/0000-0001-5868-6752
Guofeng Wu http://orcid.org/0000-0003-2275-6530

References
1. Du, C. W., and Zhou, J. M. (2011) Application of infrared photoacoustic spectroscopy in soil
analysis. Appl. Spectrosc. Rev. 46(5): 405–422. doi:10.1080/05704928.2011.570837.
2. Foy, C. D., Chaney, R. L., and White, M. C. (1978) The physiology of metal toxicity in plants.
Annu. Rev. Plant Physiol. 29: 511–566. doi:10.1146/annurev.pp.29.060178.002455.
3. Khan, S., Cao, Q., Zheng, Y. M., Huang, Y. Z., and Zhu, Y. G. (2008) Health risks of heavy metals
in contaminated soils and food crops irrigated with wastewater in Beijing, China. Environ. Pol-
lut. 152: 686–692. doi:10.1016/j.envpol.2007.06.056.
4. Horta, A., Malone, B., Stockmann, U., Minasny, B., Bishop, T. F. A., McBratney, A. B., Pallasser,
R., and Pozza, L. (2015) Potential of integrated field spectroscopy and spatial analysis for
enhanced assessment of soil contamination: A prospective review. Geoderma. 241–242: 180–
209. doi:10.1016/j.geoderma.2014.11.024.
5. Liu, Y. L., Cheng, W., and Liu, X. J. (2013) China’s food security soiled by contamination. Science
339(6126): 1382–1383. doi:10.1126/science.339.6126.1382-b.
6. Wei, B., and Yang, L. (2010) A review of heavy metal contamination in urban soils, urban road
dusts and agricultural soils from China. Microchemical Journal 94(2): 99–107. doi:10.1016/j.
microc.2009.09.014.
7. Butler, O. T., Cairns, W. R., Cook, J. M., and Davidson, C. M. (2012) Atomic spectrometry
update. Environmental analysis. J. Anal. Atom. Spectrom. 27(2): 187–221. doi:10.1039/
C1JA90057A.
8. Tuzen, M. (2003) Determination of heavy metals in soil, mushroom and plant samples by atomic
absorption spectrometry. Microchem. J. 74(3): 289–297. doi:10.1016/S0026-265X(03)00035-3.
9. Von Steiger, B., Webster, R., Schulin, R., and Lehmann, R. (1996) Mapping heavy metals in pol-
luted soil by disjunctive Kriging. Environ. Pollut. 94(2): 205–215. doi:10.1016/S0269-7491(96)
00060-7.
10. Jenny, H. (1941) Factors of soil formation – A system of quantitative pedology. McGraw Hill Book
Co, New York.
11. Sun, X. L., Zhao, Y. G., Liu, F., Wang, D. C., and Liang, C. P. (2013) Digital soil mapping and
advance in research (in Chinese). Chinese J. Soil Sci. 44(3): 752–759.
12. Brevik, E. C., Calzolari, C., Miller, B. A., Pereira, P., Kabala, C., Baumgarten, A., and Jordan, A.
(2016) Soil mapping, classification, and pedologic modeling: History and future directions.
Geoderma. 264: 256–274. doi:10.1016/j.geoderma.2015.05.017.
13. John Lu, and Z. Q. (2010) The elements of statistical learning: Data mining, inference and pre-
diction. J. R. Stat. Soc. A Stat. 173(3): 693–694. doi:10.1111/j.1467-985X.2010.00646_6.x.
14. McBratney, A. B., Mendonca Santos, M. L., and Minasny, B. (2003) On digital soil mapping.
Geoderma. 117: 3–52. doi:10.1016/S0016-7061(03)00223-4.
15. Mulder, V. L., De Bruin, S., Schaepman, M. E., and Mayr, T. R. (2011) The use of remote
sensing in soil and terrain mapping – A review. Geoderma. 162(1): 1–19. doi:10.1016/j.
geoderma.2010.12.018.
18 T. SHI ET AL.

16. Kalnicky, D. J., and Singhvi, R. (2001) Field portable XRF analysis of environmental samples.
Journal of Hazardous Materials 83: 93–122. doi:10.1016/S0304-3894(00)00330-7.
17. Bertin, E. P. (1975) Principles and practice of X-ray spectrometric analysis, 2nd Edition. Plenum
Press, New York.
18. Hou, X. D., He, Y. H., and Jones, B. T. (2004) Recent advances in portable X-ray fluorescence
spectrometry. Applied Spectroscopy Reviews 39(1): 1–25. doi:10.1081/ASR-120028867.
19. Shefsky, S. (1997) Comparing field portable X-ray fluorescence (XRF) to laboratory analysis of
heavy metals in soil. Paper presented at the International Symposium of Field Screening Methods
for Hazardous Wastes and Toxic Chemicals, Las Vegas, NV, pp. 29–31.
20. Stenberg, B., Viscarra Rossel, R. A., Mouazen, A. M., and Wetterlind, J. (2010) Visible and near
infrared spectroscopy in soil science. In Advances in Agronomy, Donald L. Sparks, Ed., Academic
Press, Burlington, MA, Vol. 107, pp. 163–215.
21. 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. doi:10.1016/j.
geoderma.2005.03.007.
22. Miller, C. E. (2001) Chemical principles of near-infrared technology. The American Association of
Cereal Chemists Inc., St. Paul, MN.
23. Shi, T. Z., Chen, Y. Y., Liu, Y. L., and Wu, G. F. (2014) Visible and near-infrared reflectance
spectroscopy – An alternative for monitoring soil contamination by heavy metal. Journal of Haz-
ardous Materials 265: 166–176. doi:10.1016/j.jhazmat.2013.11.059.
24. 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. European Journal of Soil Science 62: 598–606. doi:10.1111/j.1365-
2389.2011.01369.x.
25. Shi, T. Z., Chen, Y. Y., Liu, H. Z., Wang, J. J., and Wu, G. F. (2014) Soil organic carbon content
estimation with laboratory-based visible-near-infrared reflectance spectroscopy: Feature selec-
tion. Applied Spectroscopy 68: 831–837. doi:10.1366/13-07294.
26. Soriano-Disla, J. M., Janik, L. J., Viscarra Rossel, R. A., Macdonald, L. M., and McLaughlin, M. J.
(2014) The performance of visible, near-, and mid-infrared reflectance spectroscopy for predic-
tion of soil physical, chemical, and biological properties. Applied Spectroscopy Reviews 49: 139–
186. doi:10.1080/05704928.2013.811081.
27. Viscarra Rossel, R. A., Behrens, T., Ben-Dor, E., Brown, D. J., Dematte, J. A. M., Shepherd, K. D.,
Shi, Z., Stenberg, B., Stevens, A., Adamchuk, V., Aichi, H., Barthes, B. G., Bartholomeus, H. M.,
Bayer, A. D., Bernoux, M., Bottcher, K., Brodsky, L., Du, C. W., Chappell, A., Fouad, Y., Genot,
V., Gomez, C., Grunwald, S., Gubler, A., Guerrero, C., Hedley, C. B., Knadel, M., Morras, H. J.
M., Nocita, M., Ramirez-Lopez, L., Roudier, P., Rufasto Campos, E. M., Sanborn, P., Sellitto, V.
M., Sudduth, K. A., Rawlins, B. G., Walter, C., Winowiecki, L. A., Hong, S. Y., and Ji, W. (2016)
A global spectral library to characterize the world’s soil. Earth-Science Reviews 155: 198–230.
doi:10.1016/j.earscirev.2016.01.012.
28. 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. Australian Journal of Soil Research 46: 1–16.
doi:10.1071/SR07099.
29. Guerrero, C., Zornoza, R., Gomez, I., and Mataix-Beneyto, J. (2010) Spiking of NIR regional
models using samples from targe sites: Effect of model size on prediction accuracy. Geoderma
158: 66–77. doi:10.1016/j.geoderma.2009.12.021.
30. Wetterlind, J., and Stenberg, B. (2010) Near-infrared spectroscopy for within-field soil character-
ization: Small local calibrations compared with national libraries spiked with local samples.
European Journal of Soil Science 61: 823–843. doi:10.1111/j.1365-2389.2010.01283.x.
31. Ji, W., Li, S., Chen, S., Shi, Z., Viscarra Rossel, R. A., and Mouazen, A. M. (2016) Prediction of
soil attributes using the Chinese soil spectral library and standardized spectra recorded at field
conditions. Soil & Tillage Research 155: 492–500. doi:10.1016/j.still.2015.06.004.
32. Igne, B., Reeves, J. B., McCarty, G., and Hively, W. D. (2010) Evaluation of spectral pretreat-
ments, partial least squares, least squares support vector machines and locally weighted
APPLIED SPECTROSCOPY REVIEWS 19

regression for quantitative spectroscopic analysis of soils. Journal of Near Infrared Spectroscopy
18(3): 167–176. doi:10.1255/jnirs.883.
33. Goge, F., Joffre, R., Jolivet, C., Ross, I., and Ranjard, L. (2012) Optimization criteria in
sample selection step of local regression for quantitative analysis of large soil NIRS data-
base. Chemometrics and Intelligent Laboratory Systems 110(1): 168–176. doi:10.1016/j.
chemolab.2011.11.003.
34. Shi, Z., Ji, W., Viscarra Rossel, R. A., Chen, S., and Zhou, Y. (2015) Prediction of soil organic
matter using a spatially constrained local partial least squares regression and the Chinese vis-
NIR spectral library. European Journal of Soil Science 66(4): 679–687. doi:10.1111/ejss.12272.
35. Li, S., Shi, Z., Chen, S., Ji, W., Zhou, L., Yu, W., and Webster, R. (2015) In situ measurements of
organic carbon in soil profiles using vis-NIR spectroscopy on the Qinghai–Tibet Plateau. Envi-
ronmental Science & Technology 49: 4980–4987. doi:10.1021/es504272x.
36. Viscarra Rossel, R. A., Lobsey, C. R., Sharman, C., Flick, P., and McLachlan, G. (2017) Novel soil
profile sensing to monitor organic C stocks and condition. Environmental Science & Technology
51: 5630–5641. doi:10.1021/acs.est.7b00889.
37. Choe, E., van der Meer, F., van Ruitenbeek, F., van der Werff, H., de Smeth, B., and Kim, K. W.
(2008) Mapping of heavy metal pollution in stream sediments using combined geochemistry,
field spectroscopy, and hyperspectral remote sensing: A case study of the Rodalquilar mining
area, SE Spain. Remote Sensing of Environment 112: 3222–3233. doi:10.1016/j.rse.2008.03.017.
38. Kooistra, L., Wehrens, R., Leuven, R. S. E. W., and Buydens, L. M. C. (2001) Possibilities of visi-
ble-near-infrared spectroscopy for the assessment of soil contamination in river floodplains.
Analytica Chimica Acta 446: 97–105. doi:10.1016/S0003-2670(01)01265-X.
39. Ren, H. Y., Zhuang, D. F., Singh, A. N., Pan, J. J., Qiu, D. S., and Shi, R. H. (2009) Estimation of
As and Cu contamination in agricultural soils around a mining area by reflectance spectroscopy:
A case study. Pedosphere 19(6): 719–726. doi:10.1016/S1002-0160(09)60167-3.
40. Siebielec, G., McCarty, G. W., Stuczynski, T. I., and Reeves, J. B. (2004) Near- and mid-infrared
diffuse reflectance spectroscopy for measuring soil metal content. Journal of Environmental
Quality 33(6): 2056–2069. doi:10.2134/jeq2004.2056.
41. Vohland, M., Bossung, C., and Frund, H. C. (2009) A spectroscopic approach to assess trace-
heavy metal contents in contaminated floodplain soils via spectrally active soil components. J.
Plant Nutr. Soil Sci. 172: 201–209. doi:10.1002/jpln.200700087.
42. Wang, J. J., Cui, L. J., Gao, W. X., Shi, T. Z., Chen, Y. Y., and Gao, Y. (2014) Prediction of low
heavy metal concentrations in agricultural soils using visible and near-infrared reflectance spec-
troscopy. Geoderma 216: 1–9. doi:10.1016/j.geoderma.2013.10.024.
43. Wu, Y. Z., Chen, J., Tian, Q. J., and Qin, Z. H. (2005) Possibilities of reflectance spectroscopy for
the assessment of contaminant elements in suburban soils. Appl. Geochem. 20: 1051–1059.
doi:10.1016/j.apgeochem.2005.01.009.
44. Wu, Y. Z., Chen, J., Ji, J. F., Gong, P., Liao, Q. L., Tian, Q. J., and Ma, H. R. (2007) A mechanism
study of reflectance spectroscopy for investigating heavy metals in soil. Soil Sci. Soc. Am. J. 71:
918–926. doi:10.2136/sssaj2006.0285.
45. Wu, C. Y., Jacobson, A. R., Laba, M., Kim, B., and Baveye, P. C. (2010) Surrogate calibrations and
the diffuse-reflectance near-infrared sensing of trace metal content in soils. Water, Air, and Soil
Pollution 209: 377–390. doi:10.1007/s11270-009-0206-6.
46. Baveye, P. C., and Laba, M. (2015) Visible and near-infrared reflectance spectroscopy is of lim-
ited practical use to monitor soil contamination by heavy metals. Journal of Hazardous Materials
285: 137–139. doi:10.1016/j.jhazmat.2014.11.043.
47. Jiang, Q., Chen, Y. Y., Guo, L., Fei, T., and Kun, Q. (2016) Estimating soil organic carbon of
cropland soil at different levels of soil moisture using VIS-NIR spectroscopy. Remote Sensing 8:
755. doi:10.3390/rs8090755.
48. Liu, Y. L., Jiang, Q., Shi, T. Z., Fei, T., Wang, J. J., Liu, G. L., and Chen, Y. Y. (2014) Prediction of
total nitrogen in cropland soil at different levels of soil moisture with Vis/NIR spectroscopy.
Acta Agriculturae Scandinavica, Section B- Soil & Plant Science 64(3): 267–281.
49. Kemper, T., and Sommer, S. (2004) Use of airborne hyperspectral data to estimate residual heavy
metal contamination and acidification potential in the Guadiamar floodplain Andalusa, Spain
20 T. SHI ET AL.

after the Aznacollar mining accident. International Society for Optics and Photonics, Remote
Sensing 5574(11): 224–234.
50. Wu, Y. Z., Zhang, X., Liao, Q. L., and Ji, J. F. (2011) Can contamination elements in soils be
assessed by remote sensing technology: a case study with simulated data. Soil Sci. 176: 196.
doi:10.1097/SS.0b013e3182114717.
51. Wu, Y. Z., Chen, J., Ji, J. F., Tian, Q. J., and Wu, X. M. (2005) Feasibility of reflectance spectros-
copy for the assessment of soil mercury contamination. Environmental Science & Technology 39:
873–878. doi:10.1021/es0492642.
52. Gerighausen, H., Menz, G., and Kaufmann, H. (2012) Spatially explicit estimation of clay and
organic carbon content in agricultural soils using multi-annual imaging spectroscopy data.
Applied & Environmental Soil Science 1: 629–636.
53. Diek, S., Schaepman, M. E., and Jong, R. D. (2016) Creating multi-temporal composites of
airborne imaging spectroscopy data in support of digital soil mapping. Remote Sensing 8(11):
906–933.
54. Wilford, J., de Caritat, P., and Bui, E. (2016) Predictive geochemical mapping using environmen-
tal correlation. Applied Geochemistry 66: 275–288. doi:10.1016/j.apgeochem.2015.08.012.
55. Lado, L. R., Hengl, T., and Reuter, H. I. (2008) Heavy metals in European soils: A geostatistical
analysis of the FOREGS Geochemical database. Geoderma 148(2): 189–199. doi:10.1016/j.
geoderma.2008.09.020.
56. Peng, Y., Bou Kheir, R., Adhikari, K., Malinowski, R., Greve, M. B., Knadel, M., and Greve, M. H.
(2016) Digital mapping of toxic metals in Qatari soils using remote sensing and ancillary data.
Remote Sensing 8(12): 1003. doi:10.3390/rs8121003.
57. Farr, T. G., Rosen, P. A., Garo, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M.,
Rodriguez, E., Roth, L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner, M., Oskin, M.,
Burbank, D., and Alsdorf, D. (2007) The shuttle radar topograhpy mission. Reviews of Geophys-
ics 45(2): RG2004. doi:10.1029/2005RG000183.
58. Liu, X. Y. (2008) Airborne LiDAR for DEM generation: Some critical issues. Progress in Physical
Geography 32(1): 31–49. doi:10.1177/0309133308089496.
59. Hotelling, H. (1936) Relations between two sets of variates. Biometrika 28(3/4): 321–377.
doi:10.2307/2333955.
60. Webster, R. (1977) Canonical correlation in pedology: How useful? Eur. J. Soil Sci. 28(1): 196–
221. doi:10.1111/j.1365-2389.1977.tb02306.x.
61. Florinsky, I. V., Eilers, R. G., Manning, G. R., and Fuller, L. G. (2002) Prediction of soil proper-
ties by digital terrain modelling. Environmental Modelling & Software 17(3): 295–311.
doi:10.1016/S1364-8152(01)00067-6.
62. Moore, I. D., Gessler, P. E., Nielson, G. A., and Peterson, G. A. (1993) Soil attribute pre-
diction using terrain analysis. Soil Sci. Soc. Am. J. 57: 443–452. doi:10.2136/
sssaj1993.03615995005700020026x.
63. Gessler, P. E., Moore, I. D., McKenzie, N. J., and Ryan, P. J. (2007) Soil-landscape modelling and
spatial prediction of soil attributes. International Journal of Geographical Information Systems 9
(4): 421–432. doi:10.1080/02693799508902047.
64. Lane, P. W. (2002) Generalized linear models in soil science. European Journal of Soil Science 53:
241–251. doi:10.1046/j.1365-2389.2002.00440.x.
65. Bishop, T. F. A., and McBratney, A. B. (2001) A comparison of prediction methods for the crea-
tion of field-extent soil property maps. Geoderma 103(1): 149–160. doi:10.1016/S0016-7061(01)
00074-X.
66. Were, K., Bui, D. T., Dick, O. B., and Singh, B. R. (2015) A comparative assessment of support
vector regression, artificial neural networks, and random forests for predicting and mapping soil
organic carbon stocks across an Afromontane landscape. Ecological Indicators 52: 394–403.
doi:10.1016/j.ecolind.2014.12.028.
67. Behrens, T., Forster, H., Scholten, T., Steinr€ ucken, U., Spies, E. D., and Goldschmitt, M. (2005)
Digital soil mapping using artificial neural networks. Journal of Plant Nutrition and Soil Science
168(1): 21–33. doi:10.1002/jpln.200421414.
APPLIED SPECTROSCOPY REVIEWS 21

68. McKenzie, N. J., and Ryan, P. J. (1999) Spatial prediction of soil properties using environmental
correlation. Geoderma 89(1–2): 67–94. doi:10.1016/S0016-7061(98)00137-2.
69. Matheron, G. (1963) Principles of geostatistics. Economic Geology 58(8): 1246–1266.
doi:10.2113/gsecongeo.58.8.1246.
70. Krige, D. G. (1966) Two-dimensional weighted moving average trend surfaces for ore-evalua-
tion. Journal of the South African Institute of Mining and Metallurgy 66: 13–38.
71. Scull, P., Franklin, J., Chadwick, O. A., and McArthur, D. (2003) Predictive soil mapping: A
review. Progress in Physical Geography 27(2): 171–197. doi:10.1191/0309133303pp366ra.
72. Burgess, T., and Webster, R. (1980) Optimal interpolation and isarithmic mapping of soil prop-
erties. European Journal of Soil Science 31(2): 315–341. doi:10.1111/j.1365-2389.1980.tb02084.x.
73. McBratney, A. B., Odeh, I. O. A., Bishop, T. F. A., Dunbar, M. S., and Shatar, T. M. (2000) An
overview of pedometric techniques for use in soil survey. Geoderma 97: 293–327. doi:10.1016/
S0016-7061(00)00043-4.
74. Xie, Y., Chen, T., Lei, M., Yang, J., Guo, Q., Song, B., and Zhou, X. (2011) Spatial distribution of
soil heavy metal pollution estimated by different interpolation methods: Accuracy and uncer-
tainty analysis. Chemosphere 82: 468–476. doi:10.1016/j.chemosphere.2010.09.053.
75. Li, Y., Li, C. K., Tao, J. J., and Wang, L. D. (2011) Study on spatial distribution of soil heavy met-
als in Huizhou City based on BP-ANN modeling and GIS. Procedia Environmental Science 10:
1953–1960. doi:10.1016/j.proenv.2011.09.306.
76. Yasrebi, J., Saffari, M., Fathi, H., Karimian, N., Moazallahi, M., and Gazni, R. (2009) Evaluation
and comparison of ordinary kriging and inverse distance weighting methods for prediction of spa-
tial variablity of soil soil chemical parameters. Research Journal of Biological Science 4(1): 93–102.
77. Webster, R., and Oliver, M. A. (1992) Sample adequately to estimate variograms of soil proper-
ties. European Journal of Soil Science 43(1): 177–192. doi:10.1111/j.1365-2389.1992.tb00128.x.
78. Laslett, G. M., McBratney, A. B., Pahl, P., and Hutchinson, M. F. (1987) Comparison of several
spatial prediction methods for soil PH. European Journal of Soil Science 38(2): 325–341.
doi:10.1111/j.1365-2389.1987.tb02148.x.
79. Hengl, T., Heuvelink, G. B. M., and Stein, A. (2003) Comparison of kriging with external drift
and regression-kriging. Technical note, ITC 51: 1–17.
80. Webster, R. (1994) The development of pedometrics. Geoderma 62(1–3): 1–15. doi:10.1016/
0016-7061(94)90024-8.
81. Goovaerts, P. (1997) Geostatistics for natural resources evaluation. Oxford University Press,
London.
82. Odeh, I. O. A., and McBratney, A. B. (2000) Using AVHRR images for spatial prediction of clay
content in the lower Namoi Valley of eastern Australia. Geoderma 97(3): 237–254. doi:10.1016/
S0016-7061(00)00041-0.
83. Odeh, I. O. A., McBratney, A. B., and Chittleborough, D. J. (1995) Further results on prediction
of soil properties from terrain attributes: Heterotopic cokriging and regression-kriging.
Geoderma 67(3–4): 215–226. doi:10.1016/0016-7061(95)00007-B.
84. Hengl, T., Heuvelink, G. B. M., and Stein, A. (2004) A generic frame work for spatial prediction
of soil variables based on regression-kriging. Geoderma 120(1): 75–93. doi:10.1016/j.
geoderma.2003.08.018.
85. Odeh, I. O. A., McBratney, A. B., and Chittleborough, D. J. (1994) Spatial prediction of soil prop-
erties from landform attributes derived from a digital elevation model. Geoderma 63: 197–214.
doi:10.1016/0016-7061(94)90063-9.
86. Chen, T., Chang, Q. R., Clevers, J. G. P. W., and Kooistra, L. (2016) Identification of soil heavy
metal sources and improvement in spatial mapping based on soil spectral information: A case
study in northwest China. Science of the Total Environment 565(15): 155–164. doi:10.1016/j.
scitotenv.2016.04.163.
87. Fotheringham, A. S., Brunsdon, C., and Charlton, M. (2003) Geographically weighted regression:
The analysis of spatially varying relationships. Wiley, Hoboken, NJ.
88. Zhang, C., Tang, Y., Xu, X., and Kiely, G. (2011) Towards spatial geochemical modelling: Use of
geographically weighted regression for mapping soil organic carbon contents in Ireland. Appl.
Geochem. 26(7): 1239–1248. doi:10.1016/j.apgeochem.2011.04.014.
22 T. SHI ET AL.

89. Harris, P., Fotheringham, A., Crespo, R., and Charlton, M. (2010) The use of geographically
weighted regression for spatial prediction: An evaluation of models using simulated data sets.
Math. Geosci. 42(6): 657–680. doi:10.1007/s11004-010-9284-7.
90. Kumar, S., Lal, R., and Liu, D. (2012) A geographically weighted regression kriging approach for
mapping soil organic carbon stock. Geoderma 189: 627–634. doi:10.1016/j.
geoderma.2012.05.022.
91. Liu, Y. L., Guo, L., Jiang, Q., Zhang, H., and Chen, Y. Y. (2015) Comparing geospatial techniques
to predict SOC stocks. Soil Tillage Res. 148: 46–58. doi:10.1016/j.still.2014.12.002.
92. Guo, L., Zhao, C., Zhang, H., Chen, Y. Y., Linderman, M., Zhang, Q., and Liu, Y. L. (2017) Com-
parison of spatial and non-spatial models for predicting soil carbon content based on visible and
near-infrared spectral technology. Geoderma 285: 280–292. doi:10.1016/j.geoderma.2016.10.010.
93. Han, C., Wang, L., Gong, Z., and Xu, H. (2005) Chemical forms of soil heavy metals and their
environmental significance (in Chinese). Chinese Journal of Ecology 24(12): 1499–1502.
94. Tessier, A., Campbell, P. G. C., and Bisson, M. (1979) Sequential extraction procedure for the
speciation of particulate trace metals. Analytical Chemistry 51(7): 844–851. doi:10.1021/
ac50043a017.
95. Singh, A. K., and Benerjee, D. K. (1999) Grain size and geochemical partitioning of heavy metals
in sediments of the Damodar River – A tributary of the lower Ganga, India. Environmental Geol-
ogy 39(1): 91–98. doi:10.1007/s002540050439.
96. Zhou, Q. X., and Huang, G. H. (2001) Environmental biochemistry and global environmental
changes (in Chinese). Science Press, Beijing, China.
97. Tian, J., Wang, L., Li, X., Gong, H., Shi, C., Zhong, R., and Liu, X. (2017) Comparison of UAV
and WorldView-2 imagery for mapping leaf area index of mangrove forest. International Journal
of Applied Earth Observation and Geoinformation 61: 22–31. doi:10.1016/j.jag.2017.05.002.
98. Zarco-Tejada, P. J., Gonzalez-Dugo, V., and Berni, J. A. J. (2012) Fluorescence, temperature and
narrow-band indices acquired from a UAV platform for water stress detection using a micro-
hyperspectral imager and a thermal camera. Remote Sens. Environ. 117: 322–337. doi:10.1016/j.
rse.2011.10.007.
99. Zarco-Tejada, P. J., Guillen-Climent, M. L., Hernandez-Clemente, R., Catalina, A., Gonzalez, M.
R., and Martin, P. (2013) Estimating leaf carotenoid content in vineyards using high resolution
hyperspectral imagery acquired from an unmanned aerial vehicle (UAV). Agric. For. Meteorol.
171: 281–294. doi:10.1016/j.agrformet.2012.12.013.
100. Guo, G., Wu, F., Xie, F., and Zhang, R. (2012) Spatial distribution and pollution assessment of
heavy metals in urban soils from southwest China. Journal of Environmental Sciences 24(3):
410–418. doi:10.1016/S1001-0742(11)60762-6.
101. Maas, S., Scheifler, R., Benslama, M., Crini, N., Lucot, E., Brahmia, Z., Benyacoub, S., and Girau-
doux, P. (2010) Spatial distribution of heavy metal concentrations in urban, suburban and agri-
cultural soils in a Mediterranean city of Algeria. Environmental Pollution 158: 2294–2301.
doi:10.1016/j.envpol.2010.02.001.
102. Navas, A., and Machin, J. (2002) Spatial distribution of heavy metals and arsenic in soils of
Aragon (northeast Spain): Controlling factors and environmental implications. Applied
Geochemistry 17: 961–973. doi:10.1016/S0883-2927(02)00006-9.
103. Imperato, M., Adamo, P., Naimo, D., Arienzo, M., Stanzione, D., and Violante, P. (2003) Spatial
distribution of heavy metals in urban soils of Naples city (Italy). Environmental Pollution 124:
247–256. doi:10.1016/S0269-7491(02)00478-5.
104. Zhang, H. H., Chen, J. J., Zhu, L., Li, F. B., Wu, Z. F., Yu, W. M., and Liu, J. M. (2011) Spatial pat-
terns and variation of soil cadmium in Guangdong Province, China. Journal of Geochemical
Exploration 109: 86–91. doi:10.1016/j.gexplo.2010.10.014.
105. Ordonez, A., Loredo, J., De Miguel, E., and Charlesworth, S. (2003) Distribution of heavy metals
in the street dusts and soil of an industrial city in northern Spain. Arch. Environ. Contam. Toxi-
col. 44: 160–170. doi:10.1007/s00244-002-2005-6.
106. Martinez-Garcia, M. J., Moreno-Grau, S., Martinez-Garcia, J. J., Moreno, J., Bayo, J., Guillen
Perez, J. J., and Moreno-Clavel, J. (2001) Distribution of the metals lead, cadmium, copper, and
APPLIED SPECTROSCOPY REVIEWS 23

zinc in the top soil of Cartagena, Spain. Water, Air, and Soil Pollution 131: 329–347. doi:10.1023/
A:1011916218755.
107. McGrath, D., Zhang, C., and Carton, O. T. (2004) Geostatistical analyses and hazard assessment
on soil lead in Silvermines area, Ireland. Environmental Pollution 127: 239–248. doi:10.1016/j.
envpol.2003.07.002.
108. Saby, N., Arrouays, D., Boulonne, L., Jolivet, C., and Pochot, A. (2006) Geostatistical assessment
of Pb in soil around Paris, France. Science of the Total Environment 367: 212–221. doi:10.1016/j.
scitotenv.2005.11.028.
109. Romic, M., and Romic, D. (2003) Heavy metals distribution in agricultural topsoils in urban
area. Environmental Geology 43: 795–805.
110. Martin, J. A. R., Arias, M. L., and Corbi, J. M. G. (2006) Heavy metals contents in agricultural
topsoils in the Ebro basin (Spain). Application of the multivariate geostatistical methods to study
spatial variations. Environmental Pollution 144: 1001–1012. doi:10.1016/j.envpol.2006.01.045.
111. Lee, C. S., Li, X. D., Shi, W. Z., Cheung, S. C., and Thornton, I. (2006) Metal contamination in
urban, suburban, and country park soils of Hongkong: A study based on GIS and multivariate
statistics. Science of the Total Environment 356: 45–61. doi:10.1016/j.scitotenv.2005.03.024.
112. Sun, C. Y., Liu, J., Wang, Y., Sun, L., and Yu, H. (2013) Multivariate and geostatistical analyses of
the spatial distribution and sources of heavy metals in agricultural soil in Dehui, Northeast
China. Chemosphere 92: 517–523. doi:10.1016/j.chemosphere.2013.02.063.
113. Facchinelli, A., Sacchi, E., and Mallen, L. (2001) Multivariate statistical and GIS-based approach
to identify heavy metal sources in soils. Environmental Pollution 114(3): 313–324. doi:10.1016/
S0269-7491(00)00243-8.
114. Li, X. D., Lee, S. L., Wong, S., Shi, W. Z., and Thornton, I. (2004) The study of metal contamina-
tion in urban soils of Hong Kong using a GIS-based approach. Environmental Pollution 129:
113–124. doi:10.1016/j.envpol.2003.09.030.
115. Bou Kheir, R., Shomar, B., Greve, M. B., and Greve, M. H. (2014) On the quantitative relation-
ships between environmental parameters and heavy metals pollution in Mediterranean soils
using GIS regression-trees: The case study of Lebanon. Journal of Geochemical Exploration 147:
250–259. doi:10.1016/j.gexplo.2014.05.015.
116. Chen, T., Chang, Q. R., Clevers, J. G. P. W., and Kooistra, L. (2015) Rapid identification of soil
cadmium pollution risk at regional scale based on visible and near-infrared spectroscopy. Envi-
ronmental Pollution 206: 217–226. doi:10.1016/j.envpol.2015.07.009.
117. Zhang, C. (2006) Using multivariate analyses and GIS to identify pollutants and their spatial pat-
terns in urban soils in Galway, Ireland. Environmental Pollution 142(3): 501–511. doi:10.1016/j.
envpol.2005.10.028.

Вам также может понравиться