Agriculture Reference
In-Depth Information
weather on additional growth factors such as the visible radiation and the temperature
should also be mentioned.
It can be expected that the prediction of site-specifi c yields by means of maps
from previous harvests is better in irrigated areas, since here the soil water supply
varies less.
13.4
From Properties to Treatment - Management Zones
Map overlay can be used to create
management zones
. These can be defi ned as
areas within a fi eld to which a particular common treatment may be applied. The
basis of management zones are generally
temporally stable soil properties
such as
electrical conductivity, elevation, slope and organic matter content. The objectives
can be
e.g.
the control of cultivating, sowing and fertilizing operations. Contrary to
classical on-the-go control in real-time that might be based on just one soil- or crop
property, management zones always rely on several properties simultaneously with
the emphasis on soil attributes. Crop properties are less suited as components for
management zones as a result of their more transient character. However, site-
specifi c information about past crop yields might be included.
The preparatory steps for generating maps of individual properties have been
dealt with in Sect.
13.3
. In a
treatment map
, several properties that have been indi-
vidually mapped must be merged to singular site-specifi c signals. With the proper-
ties mentioned, this is generally not possible by applying simple mathematical
procedures. Because these properties differ substantially and hence do not allow for
this. Instead, the generating of a treatment map or a management zone for a fi eld
requires applying an algorithm. Such an algorithm consists of a set of steps in simi-
lar way as outlined in Sect.
7.2.2.1
for the control of the depth in primary cultiva-
tion. Essential is that within such an algorithm the georeferenced properties from
the individual maps are sorted into groups that are called clusters.
The processing of the grouped property signals includes a
cluster analysis
,
which is a classifi cation in such a way that association between the signals is maxi-
mal if these belong to the same group and minimal otherwise. The clustering aim is
to obtain a high internal homogeneity within the groups combined with large exter-
nal heterogeneity between the means of the groups. The cluster analysis thus simply
discovers structures in the data without explaining why these exist.
For further details to the procedures in the delineation of management zones see
Fridgen et al. (
2004
), Schepers et al. (
2004
), Taylor et al. (
2007
) and Whelan and
Taylor (
2010
).
The number of management zones that are created per fi eld can be preset in a
computer program, however, always is less than 8. The example in Fig.
13.5
shows
just two management zones. A similarly high resolution as with on-the-go opera-
tions that are based on single properties cannot be obtained. The loss in resolution
that occurs may be due to the collating process and to limitations in the algorithm.
However, present state of the art is that the same management zones are used for
