Environmental Engineering Reference
In-Depth Information
which variables are most important for explaining the variation in agronomic
soil properties (i.e. the complete set). Other variables in the complete set will be
correlated, or show less regional spatial variation. The following three principal
component analyses were performed:
-
Only European Soil Bureau (ESB) data (SGDBE and PTRDB data);
-
Only continuous variables;
-
Combination of all of variables.
A principal component analysis is most effective on one type of variables.
A mixture of continuous variables and classes is problematic. Furthermore, a
restriction in using the ESDB data is that geographical (country) borders have
large influences on the class (value) of certain soil variables. The interpretation of
certain soil variables differs between countries resulting in a heterogeneous
database. Therefore, the ESB data are difficult to use for the determination of the
most important soil factor that explains the most variation in agronomic properties
of soils within Europe.
The PCA on only the continuous data revealed that all variation in the input
variables is explained by the topsoil organic carbon content. Disadvantage is
the relatively small, but balanced dataset (OCTOP and available water holding
capacity). The PCA on all variables revealed that 95% of the input variables is
explained by the OCTOP. The overall conclusion of the PCA analysis was to
use the Topsoil Organic Carbon content as variable to differentiate between soils
in Europe.
The OCTOP is a continuous variable which has been grouped into the following
six classes (in percentage):
-
Class 1: 0.1-1.23
-
Class 2: 1.23-2.46
-
Class 3: 2.46-3.94
-
Class 4: 3.94-5.66
-
Class 5: 5.66-8.86
-
Class 6: 8.86-63.0
-
Class 9: no data or 0
The class limits of the six classes are established in such a way that each class
is covering a European land surface area of approximately the same extent.
This was achieved by using the quantiles option in ArcGIS for distribution over
six subclasses. Note that in the statistics the “no data” class is dealt with as a
seventh class.
Despite the fact that the statistical analysis identified the OCTOP as the most
suitable data to represent variation in soils properties in Europe, it is important
to note that this is likely caused by the limited quality and availability of other
high-resolution spatial data for agronomic soil properties. Despite strong correlations
with OCTOP, there is likely to be considerable variation in soil properties within
each class e.g. caused by variation in soil parent material and soil structure which
could affect issues such as rooting depth.
