Application of near-infrared reflectance for quantitative assessment of soil properties

Application of near infrared reflectance for quantitative assessment of soil properties The Egyptian Journal of Remote Sensing and Space Sciences xxx (2017) xxx–xxx Contents lists available at Science[.]

Review Article

Application of near-infrared reflectance for quantitative assessment of

soil properties

National Authority for Remote Sensing and Space Sciences (NARSS), Cairo, Egypt

a r t i c l e i n f o

Article history:

Received 24 July 2016

Revised 31 January 2017

Accepted 1 February 2017

Available online xxxx

Keywords:

Near infrared spectroscopy

Soil salinity

Soil moisture

Soil organic carbon

Soil surface features and soil contamination

a b s t r a c t

Beginning with a discussion of reflectance spectroscopy, this article attempts to provide a review on fun-damental concepts of reflectance spectroscopic techniques Their applications as well as exploring the role of Near-infrared reflectance spectroscopy that would be used for monitoring and mapping soil char-acteristics This technique began to be used in the second half of the 20th century for industrial purposes Moreover, this article explores the potentiality of predicting soil properties based on spectroscopic mea-surements Quantitative prediction of soil properties such as; salinity, organic carbon, soil moisture and heavy metals can be conducted using various calibration models – such models were developed depend-ing on the measured soil laboratory analyses data and soil reflectance spectra thereby resampled to satel-lite images - to predict soil properties The most common used models are stepwise multiple linear regression (SMLR), partial least squares regression (PLSR), multivariate adaptive regression splines (MARS), principal component regression (PCR) and artificial neural networks (ANN) Those methods are required to quickly and accurately measure soil characteristics at field to improve soil management and conservation at local and regional scales Visable-Near Infra Red (VIS-NIR) has been recommended as

a quick tool for mapping soil properties Furthermore, VIS-NIR reflection spectroscopy reduces the cost and time, therefore has a wonderful ability and potential use as a rapid soil analysis for both precision soil management and assessing soil quality

Ó 2017 Production and hosting by Elsevier B.V on behalf of National Authority for Remote Sensing and Space Sciences This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/

licenses/by-nc-nd/4.0/)

Contents

1 Introduction 00

2 Spectroscopy definitions and history of spectroscopy 00

3 Near-infrared reflectance spectroscopy 00

4 Predictive models 00

5 Applications of NIRS in soil sciences 00

5.1 Soil salinity 00

5.2 Soil moisture 00

5.3 Soil organic carbon 00

5.4 Clay minerals 00

5.5 Soil surface features 00

5.6 Soil contamination 00

6 Conclusion 00

References 00

http://dx.doi.org/10.1016/j.ejrs.2017.02.001

1110-9823/Ó 2017 Production and hosting by Elsevier B.V on behalf of National Authority for Remote Sensing and Space Sciences.

This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ).

Peer review under responsibility of National Authority for Remote Sensing and Space Sciences.

⇑ Corresponding author.

E-mail addresses: Salama55_55@yahoo.com , Salama55@mail.ru (E.S Mohamed).

The Egyptian Journal of Remote Sensing and Space Sciences xxx (2017) xxx–xxx

Contents lists available atScienceDirect

The Egyptian Journal of Remote Sensing and Space Sciences

j o u r n a l h o m e p a g e : w w w s c i e n c e d i r e c t c o m

1 Introduction

Soil is described as a heterogeneous system, their mechanisms

and processes are complex and difficult to be fully understood

Numerous traditional methods are used in an endeavour to

describe the relationship between different soil properties such

physical, chemical and principal soil components Consequently,

simple and accurate soil testing procedures are required in field

and laboratory Near-infrared reflectance spectroscopy (NIRS) is a

nondestructive systematic strategy for characterizing and

identify-ing soil properties Those techniques have been used since the

1960 to estimate moisture, protein, and oil in agricultural products

(Ben-Gera and Norris, 1968) During last three decades, numerous

studies illustrated that the spectral reflectance property of soil

samples in laboratory conditions, as well as field investigation of

soils’ characteristics, can be assessed where remote photography

materials occupy an increasingly significant place in the

organiza-tion of soil cover monitoring (Mohamed, 2013; Mohamed et al.,

2015; Saleh et al., 2015; Savin, 1993) Recently, this technique is

widely used in several fields as an amazing tool for evaluating such

agriculture, food, polymer pharmaceuticals and petrochemicals

Moreover, the technique, NIR method can be applied to predict soil

properties as additional (to laboratory analysis) or initial

assess-ment of soil quality (Demattê and da Silva Terra, 2014; Mateusz

Kania and Piotr gruba, 2016) Near-infrared reflectance

spec-troscopy (NIRS) has been used to predict several soil properties

such soil organic carbon, soil moisture content, soil contemenation,

soil salinity, etc Soil electrical conductivity can be detected using

visible, near infrared, or short-wave infrared spectral bands from

optical sensors to be promising for the detection of surface soil

salinity The intensive of reflectance is related to concentration of

soluble salts in salt-affected soils (Dwivedi and Rao, 1992; Khan

et al., 2005; Nield et al., 2007;Abdi et al., 2016) Many authors

sug-gested that, infrared and red channels are applicable methods to

monitor soil characteristics such iron oxides and soil moisture

are considered (Samsonova and Meshalkina, 2011; Sonia et al.,

2012andNiederberger et al., 2015) Near infrared (NIR) and

mid-infrared (MIR) ranges are promising technologies considered as a

quantitative ones that gives good results for heavy metals

concen-tration as there is a high correlation between pollutants and their

spectral indicators Reflectance spectroscopy techniques have been

used for retrieving and mapping the distribution of heavy metals

such as Pb at high accuracy (Samsonova and Meshalkina, 2011)

Many of regression models are used to estimate quantitative and

qualitative analyses of the various soil elements, based on

investi-gating the correlation between each element properties and the

observance for each selected wavelength However, the most

wide-spread regression models are partial least square regression (PLSR),

multivariate adaptive regression splines (MARS), ordinal logistic

regression, stepwise multiple linear regression (SIMR), artificial

neural networks (ANN), locally weighted regression (LWR) and

principal components regression (PCR) (Chang et al., 2001;

Ciurczack, 2001; Nawar et al., 2014; Fikrat et al., 2016; Zheng

et al., 2016)

2 Spectroscopy definitions and history of spectroscopy

Spectroscopy is the science that studies the interaction between

matter and its electromagnetic radiation (Crouch and Skoog, 2007)

Reflectance spectroscopy is the study of light as a function of

wave-length that has been reflected or scattered from a solid, liquid, or

gas This concept was expanded greatly to include any interaction

with radiative energy as a function of its wavelength or frequency

Spectroscopic data is often represented by aspectrum, a plot of the

response of interest as a function of wavelength or frequency (Herrmann and Onkelinx, 1986; Clark, 1999)

The history of spectroscopy began in the 17th century with Isaac Newton’s discovery of the with Isaac Newton’s discovery of the light nature and color basics He introduced the word ‘‘spec-trum” at first application to describe the rainbow of colors combi-nation to form white light During the early 1800s, Joseph von Fraunhofer made experimental advances with dispersive spec-trometers that enabled spectroscopy to become a more precise and quantitative scientific technique Since then, spectroscopy has played and continues to play a significant role in chemistry, physics and astronomy (Brand, 1995) As far as the development

of instrumentation and its breakthrough for industrial applications

in the second half of the 20th century were concerned, NIR pro-ceeded in technology jumps (Fig 1) In this respect, credit has lar-gely to be given to researchers in the field of agricultural science

At the same time, with few exceptions, comparatively low priority has been given to NIR spectroscopy in the chemical industry (Siesler et al., 2002) This technique recently has been developed into essential methods for scientific research and industrial quality control in a different applications such chemistry, environmental analysis, agriculture and as well as life sciences

3 Near-infrared reflectance spectroscopy The fundamental principle of VisNIR is based on the differences

in molecular characteristics, where spectral signatures of different materials are categorized based on their reflectance and absor-bance spectra The change in signatures is referred to vibrational extending and bending of atoms that arrange molecules and crys-tals Most soil components are usually observed in the mid-infrared region vibrations (2500–25,000 nm), with overtones and combinations found in the near-infrared region (400–2500 nm) (Clark, 1999; Shepherd and Walsh, 2002) The electromagnetic (EM) spectrum ranges from gamma (c) rays, at the shortest wave-lengths, to radio-waves, at the longest wavelengths (Fig 2) Most common sensing systems operate in one or several of the visible, infrared (IR) and microwave portions of the spectrum Sensor data covering those wavelengths are readily available from both satellite and airborne platforms (NASA, 2014) The energy of infra-red light corresponds to the energy requiinfra-red to cause molecular vibrations Moreover, the far-IR region (A = 4
l04 106nm) harmonize to molecular variations and the mid-IR (A = 2500

4
l04nm) corresponds to fundamental molecular vibrations, such as stretching, bending, wagging, and scissoring The energy

of near-IR light corresponds to overtones and combination bands

of fundamental molecular vibrations from the mid-IR (Drago, 1992; Workman, 1996) Vibrational spectroscopy is depending on interactions between the molecules and electronic field compo-nents of incident light in the mid- and near-IR region Such inter-actions result in absorption of light by molecules when the energy of incident light (Ep) is equal to the energy difference (AE) between the quantized energy levels of different vibrational states of the molecule (Fig 3) Their relationship can be expressed as:

where:

v is the frequency of incident light,

c is velocity of light,

A is the wavelength, and

h is Plank’s constant The energy difference, AE, is specified by chemical bonds of functional groups in the molecules A molecule must undergo a

change in dipole moment in order to absorb IR light Based on a

harmonic oscillator, the permitted energy states of a molecule

are given by:

where:v is the vibrational quantum number (v = 0, l, 2, .)

The fundamental vibration means that the transition from

v = 0 to v = l, according to the selection rule for a harmonic

oscil-lator Furthermore, if the chemical bond is too weak or the

atoms are too heavy, the fundamental vibration will occur at

very low frequency As a result, the higher overtones, in the

near-IR region, may not be detectable Therefore, the near-IR is

dominated by the overtones and the combinations of

fundamen-tal vibrations for O–H, C–H, and N–H found in mid-IR (Wetzel,

1983)

The amount of light absorbed is a function of the absorber

con-centration Based on the Beer-Lambert law, the relationship

between absorbance (A), transmittance (T), and concentration (c)

for monochromatic light can be expressed as follows;

A¼ logðl=TÞ ¼ logðIo=IÞ ¼ klc; ð3Þ

where:

Iois the intensity of the incident light,

I is the intensity of the transmitted light,

k is the molecular absorption coefficient, and

l is the path length of light through the sample

The molecular absorption coefficient, k, is the characteristic of each molecule and is dependent on the wavelength of the incident light

However, the reflectance of radiation from one type of surface material, such as soil, varies over the range of wavelengths in the electromagnetic spectrum and known as the spectral signature of the material (Fig 4)

4 Predictive models Prediction of different soil characteristics using spectral reflec-tions depends on statistical models that explain the relareflec-tionship

Fig 1 Development of near-infrared spectroscopy (Source: Siesler et al., 2002 ).

Fig 2 Electromagnetic spectrum (Source; NASA 2014).

E.S Mohamed et al / Egypt J Remote Sensing Space Sci xxx (2017) xxx–xxx 3

between them, most common used models are stepwise multiple

linear regression (SMLR), partial least squares regression (PLSR),

multivariate adaptive regression splines (MARS), principal

compo-nent regression (PCR) and artificial neural networks (ANN) SMLR is

a statistical method of regressing multiple variables while

simulta-neously removing those that aren’t important The choice of

pre-dictive variables is carried out by an automatic procedure

(Efroymson, 1960; Hocking, 1976; Draper and Smith, 1981; and

SAS, 1989) The variable that considered for addition to or

subtrac-tion from the set of explanatory variables in each step is based on a

form of a sequence of F-tests or t-tests The widely used algorithm

was first proposed byEfroymson (1960) The main types of

Step-wise multiple linear regression are forward selection, backward

elimination, and bidirectional elimination The forward selection involves starting with no variables in the model, testing the addi-tion of each variable using a chosen model fit criterion, adding the variable (if any) whose inclusion gives the most statistically significant improvement of the fit, and repeating this process until none improves the model to a statistically significant extent Back-ward elimination involves starting with all candidate variables, testing the deletion of each variable using a chosen model fit crite-rion, deleting the variable (if any) whose loss gives the most statis-tically insignificant deterioration of the model fit, and repeating this process until no further variables can be deleted without a sta-tistically significant loss of fit The bidirectional elimination is a combination of the forward selection and backward elimination

Fig 3 Stretching and bending vibrations.

Fig 4 Spectral resolution of some materials (Source: Short, 2011 ).

types The accuracy of SMLR model is measured as the actual

stan-dard error (SE) or the mean error between the predicted value and

the actual value in the hold-out sample (Mayers and Forgy, 1963)

Fig 5

PLS is a statistical method that finds a linear regression by

pro-jecting the predicted variables and the observable variables to a

new space (Tenenhaus et al., 2005; Vinzi et al., 2010) PLS

regres-sion is today most widely used in chemometrics, sensometrics,

and other related areas (Rönkkö et al., 2015)

MARS is a form of non-parametric regression analysis

(Friedman, 1991) MARS is an extension of linear models that

auto-matically models nonlinearities and interactions between

vari-ables MARS is also called EARTH in many implementations MARS consists of two phases: the forward and the backward pass The forward pass starts with a model consists of the mean of the response values and then repeatedly adds basis function in pairs

to the model At each step it finds the pair of basis functions that gives the maximum reduction in sum-of-squares residual error The two basis functions in the pair are identical except that a dif-ferent side of a mirrored hinge function is used for each function Each new basis function consists of a term already in the model multiplied by a new hinge function This process of adding terms continues until the change in residual error is too small to continue

or until the maximum number of terms is reached (Friedman, Fig 5 (a) PLSR in 2006; (b) MARS in 2006; (c) PLSR in 2012; and (d) MARS in 2012 (Source: Nawar et al., 2014 ).

E.S Mohamed et al / Egypt J Remote Sensing Space Sci xxx (2017) xxx–xxx 5

1993) The backward pass removes terms one by one, deleting the

least effective term at each step until it finds the best sub-model

MARS models are more flexible than linear regression models

PCR is a regression method that considers regressing the

depen-dent variable on a set of independepen-dent variables based on a standard

linear regression model, but uses PCA for estimating the unknown

regression coefficients in the model (Jolliffe, 1982) Instead of

regressing the dependent variable on the explanatory variables

directly, the principal components of the explanatory variables

are used as regressors PCR is some kind of a regularized procedure

The principal components with the higher variances are selected as

the regressors The major use of PCR lies in overcoming the

multi-collinearity problem which arises when two or more of the

explanatory variables are close to being collinear (Dodge, 2003)

PCR can result in dimension reduction through substantially

low-ering the effective number of parameters characterizing the

under-lying model PCR can lead to efficient prediction with the

appropriate selection of the principal components to be used for

regression

ANN is a computing system made up of a number of simple,

highly interconnected processing elements, which process

infor-mation by their dynamic state response to external inputs

(Caudill, 1987) ANNs are processing algorithms that are loosely

modeled after the neuronal structure of the mammalian cerebral

cortex but on much smaller scales Neural networks are typically

organized in layers Layers are made up of a number of

intercon-nected ‘nodes’ which contain an ‘activation function’ Patterns are

presented to the network via the ‘input layer’, which

communi-cates to one or more ‘hidden layers’ where the actual processing

is done via a system of weighted ‘connections’ The hidden layers

then link to an ‘output layer’ where the answer is the output Most

ANNs contain some form of ’learning rule’ which modifies the

weights of the connections according to the input patterns ANNs

provide an analytical alternative to conventional techniques which

are often limited by strict assumptions of normality, linearity, and

variable independence

5 Applications of NIRS in soil sciences

As mentioned above, Near-infrared reflectance spectroscopy

(NIRS) has been used to predict several soil properties There are

many authors focuses their works on predicting soil characteristics

based on Near-infrared reflectance spectroscopy, some of them

selected one parameter of soil with reflectance spectroscopy

Fur-thermore other authors predicted soil parameters as relation with

reflectance spectroscopy (Zbíra et al., 2016) Some applications will

be discussed as follows:

5.1 Soil salinity

Salinization is an overall issue that influences the physical and

chemical soil properties that leads to loss in yield efficiency

Throughout the previous two decades, remotely detected

symbol-ism has exhibited its capacity to evaluate saltiness changes at

dif-ferent scales (Metternicht and Zinck, 2008;Elnaggar and Noller,

2009) Numerous studies have illustrated the ability of Vis-NIR

reflection spectroscopy bands – from the optical sensors – for

detecting surface soil salinity Furthermore, hyperspectral data

have been used in several approaches for quantitative assessment

of soil salinity and different soil properties (Dehaan and Taylor,

2003; Farifteh et al., 2008; Feyziyev et al., 2016) It has been

sho-wen that, effective prediction of saltiness is administered by the

relationship between other soil properties, such as soil moisture

(Ben-Dor et al., 2002) For multispectral image studies, the

inclu-sion of topographic data is sometimes used to mitigate the poor

diagnostic power of the sensor and improve the classification For example, a study utilized Landsat Thematic Mapper (TM) data and Digital elevation model (DEM) obtained topographical indices for mapping soil salinity in Western Australian (Caccetta et al.,

2000) Furthermore, hyperspectral data increase the capability of remotely sensed information, thereby, can be applied more inde-pendently of other data sets The absence of spectral features of salt still makes classification difficult However, several researchers have concluded that soil salinity can be mapped based on other properties of soil as alternatives.Ben-Dor et al (2002)reported that, hyperspectral scanner data was used for mapping soil salinity, also there was a correlation between soil moisture and salinity reached (r = 0.58) in cultivated crops and was able to develop reli-able prediction equations Moreover, hyperspectral remote sensing data have been utilized to monitor soil salinity under different environmental conditions, as well as other halophyte species such

as Sea Blite and Sea Barley Grass (Dehaan and Taylor, 2003).Nawar

et al (2014)coupled MARS, PLSR and NIR soil spectra and geo-statistics to map spatial variation of soil salinity in El-Tina Plain, north Sinai, Egypt They measured electrical conductivity (ECe) data and eflectance spectra of soil samples resampled to satellite sensor’s resolution (Fig 5) The study reported good results for the prediction of soil salinity; MARS (R2 = 0.73), RMSE = 6.53, and ratio of performance to deviation (RPD) = 1.96), while PLSR model (R2 = 0.70, RMSE = 6.95, and RPD = 1.82).Moreover, the authors emphasized that MARS gives very good results for prediction of soil salinity, especially under high salinity levels Thus, it is important

to monitor and map soil salinity at an early stage to enact effective soil reclamation program that helps to lessen or prevent future increase in soil salinity Remote sensing has more informative and professional rapid assessment of soil salinity, compared with traditional methods offering more informative and professional rapid assessment techniques for monitoring and mapping soil salinity Soil salinity can be identified from remote sensing data obtained by different sensors based on visible direct indicators that refer to salt features at soil surface indicators, such as the presence

of halophytic plant

5.2 Soil moisture Previous numerous studies have shown the role of reflectance spectroscopy for monitoring soil moisture Many studies illus-trated the inverse relationship between soil moisture and spectral reflectance (Post et al., 2000; Galvao et al., 2001) Furthermore, the inverse relationship means the decrease of reflectance with the increase of soil moisture content This relationship is due to two reasons; soil particles covered with thin films of water and water

on the lattice sites of some minerals present in the soil (Stoner and Baumgardner, 1981) With the improvement of measurement tools, the change in spectral reflectance with change in soil mois-ture levels became more pronounced at longer wavelengths (>1450 nm) (Weidong et al., 2002) The same study also showed that, at higher moisture contents the trend is changed and the reflectance increased with the increasing of moisture content They determined this type of reversal to be somewhere around field capacity, while it changed for different soils, and happens before the point where water retention is saturating the reflectance signal

Bogrekci and Lee (2006)investigated the possibility of estimat-ing phosphorus by spectral reflectance under the influence of dif-ferent levels of soil moisture with difdif-ferent phosphorus (P) concentrations (0, 12.5, 62.5, 175, 375, 750, and 1000 mg kg1) using ultraviolet (UV), visible (VIS), and near-infrared (NIR) absor-bance spectroscopy (Fig 6) The authors illustrated that the mois-ture content affected the absorbance spectra, where correlation coefficient between spectra absorbance and P concentrations

showed high values within the 1982–2550 nm range In addition,

spectral signal processing by removing the moisture content effect

enhanced P prediction in soils considerably (Fig 7)

The study of any soil property is related to the understanding of

sensitive areas at the spectrum due to presence of water The

vibrational frequencies of water molecules after 2500 nm affect

the water absorption wavelengths (Baumgardner et al., 1985)

The 1450 and 1950 nm wavelengths are the absorption bands with

sharp peaks (Fig 8) The broad unordered bands are more common

in naturally occurring soils in addition, the highest significant vari-able in determining the reflectance located within a range 2080–

2320mm (Baumgardner et al., 1985andGalvao et al., 2001) The broad unordered bands are more common in naturally occurring soils Furthermore, the highest significant variable in determining the reflectance changes in the 2080–2320mm However, other studies emphasized on the importance role of reflectance

Fig 6 Average Spectral reflectance at different level of soil moisture (Source: Bogrekci and Lee, 2006 ).

Fig 7 Correlation coefficient spectra between absorbance and P concentration at different moisture contents within the 225–2550 nm range (Source: Bogrekci and Lee,

2006 ).

E.S Mohamed et al / Egypt J Remote Sensing Space Sci xxx (2017) xxx–xxx 7

spectroscopy and remote sensing to develop spectral models for

detecting soil moisture content (Ben-Dor et al., 2002; Whiting

et al., 2004)

5.3 Soil organic carbon

Soil organic carbon (SOC) is a key characteristic of soil quality

which impacts the assortment of organic compounds and physical

properties of soils (Carter, 2002) The evaluation of greenhouse gas

emissions from soils requires a precise information on the fate of

carbon and nitrogen in soils Near-infrared reflectance

spec-troscopy (NIRS) is a quick and non-damaging explanatory

proce-dure that includes diffuse-reflectance estimation in the near

infrared region (1000–2500 nm) Visible–NIR spectroscopy with

decision-tree modeling can fairly and accurately, with small to

moderate uncertainty, predict soil organic carbon (

Viscarra-Rossel and Hicks, 2015and Hu et al., 2015) Soil organic matter

(SOM) decrease the vis-NIR spectral reflectance range (520–

800 nm), especially if the SOM content is bigger than 2% (Stoner

and Baumgardner, 1981; Henderson et al., 1992) Humic acid

con-sidered the most dark pigment of SOM and reduces the spectral

reflectance over the visible to short-wave spectral range

Other-wise, fulvic acid has no influence on soil reflectance (Henderson

et al., 1992) A study of soils in Thailand using artificial neural

net-works found that vis-NIR VNIR spectrum (400–1100 nm) as a

pre-cise detector of SOM (R2= 0.86) (Daniel et al., 2003) Furthermore,

other study (Ben-Dor et al., 2002) has used hyperspectral images

for mapping SOM based on the reflectance spectra of heavy clay

soils in Israel where the root mean square of the prediction

equa-tions was (R2 m > 0.82) A support vector machine regression

(SVMR) and a successive projections algorithm (SPA) model

(SPA-SVMR model) have been used for improving the accuracy of soil

organic carbon (SOC) which has resulted from integrating the

laboratory-based visible and near-infrared (VIS/NIR, 350–

2500 nm) spectroscopy of soils (Xiaoting et al., 2014) Another

image study used digitized color aerial photography to successfully

map SOM based on two approaches The first attempt was to study

the individual pixels thereby describe the spatial distribution; the

second attempt was applying the relationship on image

classifica-tion to determine the classes units (Fig 9)

5.4 Clay minerals Soil chemistry affects clay minerals, thereby the soil develop-ment and their fertility Many of clay minerals have unique spectral reflectance at visible wavelengths and NIR-SWIR (Hunt, 1980) Silva

et al (2016) illustrated that near-infrared region can be used to predicate soil attribute with PLSR using a limited spectral region (325–1075 nm) performed poorly for sand while more promising when considering the capabilities to predict silt and clay The appli-cation of visible and part of the (400–980 nm) for clay prediction in Oxisols achieved relative good results where regression coefficients showed good relation to the spectral behavior of weathered soils in visible and near-infrared region The components of soil minerals affect the spectral reflectance of the soil through the absorption bands and overall spectral brightness Quartz is the biggest and most regular part of soils; it shows no unique absorption feature over Vis-NIR-SWIR range although it does increase the overall brightness Clay minerals have unique absorption bands that are effected by distinctive vibrational overtones, electronic and charge transfers, and conduction processes (Clark, 1999) The wavelengths around 2200 nm for the spectral characteristics of clay minerals were extracted from AVIRIS data for the identification of smectite, kaolinite and illite clay minerals (Chabrillat et al., 2002) The alter-ation phases were mapped based on absorption band position, depth and asymmetry from AVIRIS data (van-der-Meer, 2004) As vegetation obscure the target material partially by the large distinc-tive absorption features, the Mineralogical identification achieved features (Chabrillat et al., 2002) Similarly, absorption band posi-tion, depth and asymmetry have been used to map alteration phases with AVIRIS imagery (van-der-Meer, 2004) Mineralogical identification has been achieved when the target material is par-tially obscured by vegetation due the largely distinctive absorption features (Chabrillat et al., 2002) In the spectrum of hematite (an iron-oxide mineral), the strong absorption in the visible light range

is caused by ferric iron (Fe+3) In calcite, the major component of limestone, the carbonate ion (CO3) is responsible for a series of absorption bands between 1.8 and 2.4mm (mm) The most common clay minerals in soil are kaolinite and montmorillonite, these min-erals are distinct from others depending on the absorption spec-troscopy bands where the highest absorption band around 1.4mm

Fig 8 Atmospheric water absorption bands.

in wavelength, along with the weak 1.9mm band in kaolinite,

refer-ring to hydroxide ions (OH-), while the stronger 1.9mm band in

montmorillonite is affected by bound water molecules in this

hydrous clay (Fig 10) On the other hand, feldspar, the dominant

mineral in granite – shows no significant absorption features in

the vis-NIR-SWIR (Hauff et al., 1991; Masinter and Lyon, 1991)

The combination of spectroscopy reflectance data and

hyperspec-tral satellite images give remarkable results for deriving dominant

clay mineral The results from modeling dominant clay minerals by

random forests and mapping of hyperion data using Spectral

Angu-lar Mapper (SAM) illustrated the dominance of kaolinite clay

min-eral followed by montmorillonite in Madhya Pradesh India

(Fig 11) (Janaki et al., 2014)

5.5 Soil surface features The identification of surface soil features and land resources are very important for precise management in different scales The spectral signature of each soil property influenced by spatial and temporal variability of surface processes however, it is difficult to measure directly from their reflectance spectra even under con-trolled laboratory conditions (Silva and ten Caten, 2016) Soil Vis-NIR (350–2500 nm) reflectance spectra contain valuable infor-mation for predicting soil textural fractions (sand, silt, and clay content) Chemometrics techniques and multivariate calibration (PLSR) allowed researchers to extract the relevant information from the reflectance spectra and to correlate this with the soil

Fig 9 Spatial distribution of soil organic carbon (Source: Chen et al., 2000 ).

E.S Mohamed et al / Egypt J Remote Sensing Space Sci xxx (2017) xxx–xxx 9

texture fractions The same author illustrated that soil texture can

be predicted where sand content (R2 = 0.81) and clay content

(R2 = 0.80) and less satisfactory for silt content (R2 = 0.70).The

spectral signature from an image pixel is a mixture ofsurface

mate-rials affected by their chemical components The spectral

proper-ties of a single image pixel is the representation of the surface

components Each pixel retains the characteristic features of the

individual spectra from each of the component reflective materials

When the ground material- such as soil types – occupies the whole

pixel, the pixel spectra is the signatures of the ground material (Roberts et al., 1993) Saleh et al (2013) used a linear spectral unmixing analyses to discriminate different surface soil types in north sinia – Egypt by Near-infrared reflectance spectroscopy tech-niques (Fig 12) The spectra of the soil types were significantly influenced by the different surface features presented in the area The same author concluded that linear spectral unmixing is very helpful tool for identifying and mapping the different surface soil types from ETM + by discriminating the different mixture spectra

Fig 10 Spectral reflectance of different clay minerals (Source: Clark, 1999 ).

Fig 11 Dominant clay mineral in Madhya Pradesh India (Source: Janaki et al., 2014 ).