Agriculture Reference
In-Depth Information
Table 5.4
Correlation between soil properties in laboratories and ranges of reflectance
Correlation coefficients squared (r
2
)
Range of relectance
a
Visible
Soil properties
Near-infrared
Mid-infrared
From Viscarra Rossel et al. (
2006
), summary of extensive literature review
Organic carbon content
0.78
0.81
0.96
Clay content
0.71 combined
0.82
Cation-exchange-capacity
0.73 combined
0.88
From experiments of Viscarra Rossel et al. (
2006
)
Organic carbon content
0.60
0.60
0.73
Clay content
0.43
0.60
0.67
Silt content
0.31
0.41
0.49
Sand content
0.47
0.59
0.74
Cation-exchange-capacity
0.16
0.13
0.34
From experiments of McCarty et al. (
2002
)
Organic carbon content, 1.series - 0.82 0.94
Organic carbon content, 2.series - 0.98 0.98
a
The ranges are for visible reflectance 400-700 nm, for near-infrared reflectance 700-2,500 nm
and for mid-infrared reflectance 2,500-25,000 nm
this soil property makes “sense” via reflectance due to the transient situation and the
limitation to the soil surface. An essential point for soil water sensing is whether this
is done on the top-surface or along surfaces of vertical cross-sections within the soil
(see above).
It is expected that for sensing in laboratories, mid-infrared radiation will replace
the hitherto dominating near-infrared reflectance due to the more precise indica-
tions. The situation is different for sensing in fields in an on-the-go mode. Because
when using mid-infrared radiation, the soil has to be rather dry - a prerequisite that
hardly can always be met in fields. Mid-infrared radiation is absorbed very strongly
by moist soil and consequently not enough of it is reflected for sensing (Christy
2008
). Visible and near-infrared radiation is less absorbed by moist soils. This
allows measurements from moist field samples - a prerequisite of on-the-go signals
for simultaneous online control of field machinery or for mapping. An important
point is also the investment for the sensing instruments. The longer the wavelengths,
the more expensive the spectroscopic implements are. Yet technological progress is
more and more reducing the differences in investment.
Figure
5.24
shows similar results from full spectrum sensing of multiple soil
properties with processing of the data by partial least squares regression. The soil
samples came from ten fields in various regions of the Midwestern United States
and were either taken in a segmented manner from the pedogenic horizons along
vertical soil profiles
or from
top surfaces
at various sites within each field. From
the various soil properties that were included in the investigation, only the results
for the most important natural properties - organic matter and clay content - are
shown. These properties provided the best correlations between the traditional anal-
yses and the reflectance sensing in the order organic carbon, clay.
