Global Journal of Agricultural Innovation, Research & Development, 2014, 1, 17-26 17
E-ISSN: 2409-9813/14 © 2014 Avanti Publishers
Vis-NIR Spectroscopy for Determining Physical and Chemical Soil
Properties: An Application to an Area of Southern Italy
Massimo Conforti and Gabriele Buttafuoco
*
Institute for Agricultural and Forest Systems in the Mediterranean (ISAFOM), National Research Council
(CNR) of Italy, Rende (CS), Italy
Abstract: The development of rapid, accurate, cost effective methods to determine soil physical and chemical properties
is important for sustainable land management. In the last two to three decades, the interest in using visible and near
infrared (Vis-NIR) spectroscopy as an alternative method for determining soil properties has increased. To obtain reliable
predictions of soil properties, multivariate calibration techniques such as Partial Least Squares Regression (PLSR) are
commonly used to correlate the spectra with the chemical, physical and mineralogical properties of soils.
The objective of the paper was to assess the potential of Vis-NIR spectroscopy coupled with PLSR to determine soil
chemical and physical properties such as organic carbon (SOC), sand, silt, clay, and calcium carbonate (CaCO3)
contents in a sample site of southern Italy.
Spectral curves showed that the soils could be spectrally separable on the basis of chemical and physical properties.
PLSR calibration models were derived for each of the soil properties and were validated with an independent data set.
The optimum number of factors to be retained in the calibration models was determined by leave-one-out cross-
validation. The accuracy of the calibration and validation models for the different soil properties was evaluated with the
coefficient of determination (R
2
) and the root mean squared error (RMSE). The results showed that predictions were
satisfactory for all soil properties analyzed with high values of R
2
> 80.
A combination of Vis-NIR spectroscopy and multivariate statistical techniques, therefore, can be used as a rapid, low
cost and quantitative means of characterizing the soils of southern Italy.
Keywords: Soil properties, Vis-NIR reflectance spectroscopy, PLSR, Southern Italy.
1. INTRODUCTION
Soil is one of the most important natural resources
because it plays a key role in biochemical and
geochemical cycling, water partitioning (storage and
release), land protection and buffering, and energy
partitioning, all of which are essential for supporting
ecosystems [1]. Moreover, soil represents the largest
pool of carbon (C), storing approximately 1500 PgC in
the top 1 m [2, 3]; hence, even relatively small changes
in soil C storage per unit area could have a significant
impact on the global carbon balance. Therefore,
accurately quantifying soil properties and their spatial
and temporal variability is an important issue for
sustainable land management, precision agriculture
and soil mapping as well as for carbon sequestration.
However, conventional methods of estimating soil
properties, based on field or laboratory methods, are
relatively complex, time consuming, and expensive
when large numbers of soil samples need to be
analyzed [4]. To overcome this, visible and near-
infrared spectroscopy (Vis-NIR, 350 - 2500 nm) has
become, in recent decades, an important tool for
quantitative evaluation of soil properties, e.g.,
*Address correspondence to this author at the Institute for Agricultural and
Forest Systems in the Mediterranean (ISAFOM), National Research Council
(CNR) of Italy, Rende (CS), Italy; Tel: +390984841484;
Fax: +390984841497; E-mail: gabriele.buttafuoco@cnr.it
carbonate content, organic carbon (OC), total nitrogen
(N), iron (Fe) oxide minerals, soil texture [5-10].
Compared to conventional laboratory methods, the Vis-
NIR spectroscopy technique has been accepted as
rapid and cost effective, requiring minimal sample
preparation and no hazardous chemicals. It is non-
destructive and several soil properties can be
determined from a single measure [e.g. 4, 10-15]. The
Vis-NIR spectroscopy method is based on the simple
assumption that the soil reflectance in the 350-2500 nm
spectral region is a linear combination of the spectral
signatures of its various components weighted by their
abundance [16, 17]. So, small changes in physical,
chemical, mineralogical and biological soil properties
produce different spectral characteristics that can be
identified by reflectance spectroscopy [12, 18-23].
Although measuring soil Vis-NIR reflectance
requires only a few seconds, the reflectance spectra
are largely non-specific due to interference resulting
from the overlapping spectra of soil constituents that
are themselves varied and interrelated [24]. Therefore,
the physical, chemical and mineralogical properties of
soils can be correlated to reflectance spectra by
suitable multivariate calibration procedures [15, 25, 26]
such as multiple linear regression (MLR), principal
components regression (PCR), partial least-squares
regression (PLSR) and artificial neural networks (ANN)
[e.g. 6, 7, 12, 14, 21, 27-29].