Agriculture Reference
In-Depth Information
Both the calibration as well as the validation can be based on
full reflectance
spectra
or on
discrete wavebands
. In case of full spectra, the objective can be
simultaneous sensing of several soil properties in one operation. Up to now, the
results with simultaneous sensing of several soil properties via full spectra are good
or satisfying with some properties, with others they are disappointing (Table
9.2
).
Possible reasons for this have been dealt with above.
An alternative is to base the first step - the calibration procedure - on full spectra
but to extract from this information in a second step the knowledge for sensing
single soil properties by discrete wavebands of small width. With this method, the
searching of the correlations relies on a rather broad range, the validation is limited
to one or a few narrow bands. And these narrow wavebands finally are used for
sensing in site-specific applications. Because of the small bandwidths that are
involved, this method improves the chance that interferences with spectral ranges
that are affected by other soil properties can be avoided. It is probably for this rea-
son that this method of
single property sensing
by discrete wavebands is dominat-
ing up to now in practical applications. However, the narrower the wavebands are on
which the sensing relies, the more important get effective processing methods for
removing noise and for smoothing the data (Fig.
9.10
).
An interesting approach for the future might be simultaneous
hyperspectral
sensing of several properties
. The term “hyperspectral” in this respect indicates
too that only very distinct narrow wavebands are selected for every soil property and
that spectral regions in between are disregarded in order to remove the probabilities
for interferences.
9.3.2
Sensing the Lime Requirement
On-the-go sensing of the soil water pH by ion-selective electrodes in combination
with electrical conductivity recording for indicating the buffering properties of soils
and hence the lime required is state of the art (Sect.
9.2.2
). Yet this method cannot
be used to control the lime application in
real-time
because of the time delay that is
associated with sensing by ion-selective electrodes. Sensing of the lime requirement
by reflectance would remove this time problem.
Sensing the soil water pH alone by reflectance does not suffice. In order to
account for the varying
buffering properties
of soils, Viscarra Rossel et al. (
2006
)
used full spectra reflectance sensing techniques in laboratories for Australian soils
and processed the data by multivariate analyses with partial least squares regres-
sions. For the estimation of the lime needed, not only soil water pH was the subject
of spectral sensing, but several soil properties as
e.g.
cation exchange capacity,
organic carbon and texture were considered in the program as well.
The results that were thus obtained depended highly on the spectral range.
When solely visible reflectance (400-700 nm) was used, the spectroscopic results
were not reliable (Fig.
9.11
). The near-infrared reflectance (700-2,500 nm)