Agriculture Reference
In-Depth Information
obviously result from technical noise and not from the variability of the respective
soil- or crop property, should be removed. Such artefacts can be generated when
e.g. the tractor stops or starts.
• All maps must be accurately geo-encoded so that the geographic coordinates
precisely match. Distance- or time lags that are inevitable with sensing and con-
trol operations must be compensated for (see Sect. 9.4.8 ) .
• Most software programs start with the creation of a point-based map , in which
each point represents a signal. But map overlay and the corresponding data
fusion process need a grid- or block based map , for which the data around each
point are spread out in a square. When the grids or blocks are contiguous, a sur-
face map is created.
• Such a block based surface map requires to determine adjacent points by inter-
polation. Computers can easily do this automatically, however, this interpolating
should never occur beyond the range of the variogram that holds for the respec-
tive signals (see Sect. 2.3 ) . Kriging must be included in the processing, and for
this a special computer program (VESPER) from the Australian Centre for
Precision Agriculture of the University of Sydney can be helpful (Whelan et al.
2002 ; Whelan and Taylor 2010 ).
These steps are state of the art in geostatistics, but nevertheless indispensable
prerequisites for any data fusion by map overlay. The most important prerequisite
however is that the combination of properties or of signals must make sense. This
might depend on the mathematical logic that is used to merge the respective infor-
mation as with any data fusion.
There exist numerous applications for this geostatistical route from property
maps to a treatment map. An example for site-specifi c soil cultivation is outlined in
Sect. 7.2.2.1 . And a rather simple application that is oriented at environmental
control in the application of a pre-emergence herbicide named Isoproturon is pre-
sented below.
The objective was the prevention of herbicide leaching into the groundwater
(Mertens 2008 ). For this, the absorbing capacity of the soil is important. As for the
soil in Fig. 13.3 , the silt-loam on top of sandy gravel deposits has a largely varying
vertical thickness, so the leaching of the herbicide Isoproturon as a result of this
thickness was determined. Hence the site-specifi c information about the risk of
leaching could be derived from the map about the vertical thickness of the silt-loam.
This map was obtained via sensing of electrical conductivity by the deep induction
method. Though generally this method does not provide information about the reso-
lution in a vertical direction, in this case it was possible. Because there were roughly
just two vertical layers, the loess-soil on top and the sandy gravel below.
A logical procedure would be to put in addition a site-specifi c map about the local
weed infestation underneath the stack in Fig. 13.3 . In most cases, the weed infestation
has a patchy pattern. It would then be possible to see how the weed patches corre-
spond to the areas of different risks for leaching. In case the weed patches would not
fall into areas of high or very high leaching risk, the application of Isoproturon might
be acceptable if it is done in a site-specifi c mode. But if weed patches are within
Search WWH ::




Custom Search