Environmental Engineering Reference
In-Depth Information
After the selection of the soil variable that is most important from an agronomic
perspective, the EnZ is combined with the selected soil variable. Furthermore, an
Agri-mask is added to distinguish zones that are very much limited by climatic, soil
and/or other biophysical factors, while in other areas the natural factors provide
good opportunities.
The different databases have been prepared to fit the SEAMLESS spatial
data framework. If necessary they were converted from vector into raster
(1 × 1 km resolution) and transformed into the European standard projection
ETRS_1989_LAEA. The geographical coverage is limited to EU27 and Norway,
Switzerland and Balkan countries (i.e. EU27+) (see also Baruth et al. 2006a) .
Furthermore, the “empty” SEAMLESS grid cells of EnZ are filled by Euclidean
Distance routine.
Selection of Soil Variables
As it is clear that the EnZ is not representing the diversity in soil factors, it is needed
to determine the soil variable(s) that explains most of the variance from agronomic
perspective within Europe. The following soil variables were selected as important
from an agronomic perspective:
TOPsoil Organic Carbon (OCTOP) (continuous variable);
Available water holding capacity (continuous variable, PTRDB);
Rooting depth (classes, SGDBE);
Depth of gleyed horizon (classes, PTRDB);
Topsoil textural classes (classes, SGDBE);
Topsoil Cation Exchange Capacity (classes, PTRDB).
These variables come from different data sources, and have different data
properties. For example rooting depth is only available in a few classes, while
organic carbon is available as a continuous variable, showing far more local
variation. It is therefore necessary to analyze and compare these datasets, and
examine how each of them could contribute to the soil stratification, and how much
each available variable actually contribute to the explanation of soil differences.
There will be much overlap in explained spatial variation between these variables,
e.g. a soil with limitation in rooting depth will be associated with certain texture
classes, and may also have gleyed horizons. This makes that only a limited number
of data layers will be used for the stratification.
PCA was used to screen the selected variables. PCA is an effective multivariate
technique to reduce the variation of many variables into a limited number of
dimensions. The eigenvectors of the principal components explain how much
of each component is explained by each variable. In this way it is possible to detect
Search WWH ::

Custom Search