Agriculture Reference
In-Depth Information
Table 9.2
Correlations between optical reflectance and various soil properties from spectroscopic
experiments in laboratories
Coefficients of determination (r
2
)
Chang et al. (
2001
)
a
Soil property
He et al. (
2007
)
b
Lee et al. (
2009
)
c
Natural properties
Water
0.84
-
-
Organic carbon
0.60
0.92
0.80
Total N
0.85
0.88
0.53
Clay
0.67
-
0.76
Silt
0.84
-
0.79
Sand
0.82
-
0.79
Properties related to pH (extractants in brackets)
pH (water)
0.55
0.87
0.68
Ca (ammon. acetate)
0.75
-
0.80
Mg (ammon. acetate)
0.68
-
0.73
Nutrients, plant available (extractants in brackets)
K (ammon. acetate)
0.55
0.58
0.13
P (Mehlich, Olsen)
0.40
0.29
-
Cu (Mehlich)
0.25
-
-
Mn (Mehlich)
0.70
-
-
Zn (Mehlich)
0.44
-
-
The wavelength ranges used were
a
Chang et al.
2001
: 1,300-2,500 nm
b
He et al.
2007
: 350-2,500 nm
c
Lee et al.
2009
: 350-2,500 nm
In theory, soil nutrients have an optical fingerprint that is expressed in its reflectance
as well as other soil constituents have. However, it may be more difficult to detect
their fingerprints because of
interferences
. In case these interferences result from
other soil constituents, targeted processing of spectral curves might help to differen-
tiate. However, in addition to interferences from other soil properties, there may be
substantial data scattering or
signal noise
due to a too fast response, to inertial reac-
tions in vibrating instruments or many other uncontrolled factors.
A vast number of different techniques for
removing noise
is available and can be
used for online data processing such as
• computing weighted moving averages that create a new sequence of data
• smoothing and differentiating of data by simplified least squares procedures
(Savitzky-Golay filtering)
• employing first- or second derivatives of full spectra for attenuating the influence
of solar angles as well as of the viewing geometry
• using a Fourier analysis or -transformation of full spectra, which is a method of
decomposing the signals into sine waves of different frequencies
• transforming a full or partial spectrum by wavelets (= small waves). This
method is based on a frequency-time analysis, which supplements the general
frequency- amplitude approach.
