Sensing of Natural Soil Properties - Precision in Crop Farming - page 89

whatever method is used, the accurate calibration of the sensing equipment
deserves attention. The calibration is an adjusting procedure. It aims at getting the
right scale or relation between property data on the one hand and spectral signals on
the other hand. The best data basis for an accurate calibration can be provided by the
respective field, for which the sensing of properties is pending. But theoretically this
is impossible, since it implies direct converting of output to input. Hence the input
for the calibration must be supplied from another field that has a similar soil. And
since worldwide a huge variety of different soils exist, this means that a reservoir for
site-speciic spectral calibration data is needed. Brown et al. ( 2006 ) have started to
develop such spectral soil reflectance libraries.

5.3.2 Results of Surface Sensing in Laboratories

Figure 5.23 shows effects of moisture, organic matter and texture plus iron on visi-
ble and near-infrared reflectance of soils. Increasing contents of water, organic mat-
ter or clay result in decreasing reflectance. However, in the same order of soil
properties, this effect becomes smaller. As for texture, this effect is ambiguous in
the visible- and in the near-infrared range that is adjacent to it. The iron content of
the soil may be important as well.
For a discrete waveband approach , the best wavelengths (fingerprints) are
important. Within the visible and near-infrared range these are (Lee et al. 2009 ;
Mouazen et al. 2007 ; Shonk et al. 1991 ; Zhu et al. 2010 ):

for water
970 nm; 1,200 nm; 1,400 nm; 1,450 nm ; 1,820 nm; 1,940 nm ; 2,000 nm; 2,250 nm,
for organic matter
660 nm; 1,772 nm ; 1,871 nm ; 2,070 nm; 2,177 nm; 2,246 nm; 2,351 nm; 2,483 nm,
for clay
1,877 nm; 1,904 nm; 2,177 nm; 2,192 nm; 2,201 nm ; 2,220 nm; 2,492 nm and
for cation-exchange-capacity
1,772 nm; 1,805 nm; 1,877 nm ; 2,090 nm; 2,276 nm; 2,306 nm; 2,498 nm.

Key wavelengths are indicated in bold .
The wavelengths belong - with one exception - to the near-infrared range, which
is defined here as extending between 700 and 2,500 nm.
Coefficients of determination or squared correlation coefficients (r 2 ) that are
based on full spectrum sensing of numerous soil samples via reflectance in labora-
tories are listed in Table 5.4 . The respective full spectra were used to indicate vari-
ous soil properties simultaneously. Only some of the sensed soil properties are
shown. The data were processed by partial least squares regressions.
The summarized result is that the accuracy of sensing for the soil properties
listed increases in the order visible-, near-infrared- and finally mid-infrared reflec-
tance, hence with the wavelengths. In most cases, the results for organic carbon
excelled those for all other properties. For the silt- and sand content, it must be taken