Fjölrit RALA - 15.06.2004, Side 21

Multivariate statistical analysis of reference volcanic European soils
E. García-Rodeja1 A. Martínez-Cortizas1, J.C. Nóvoa1, X. Pontevedra1, and
P. Buurman2
1 Dept. Edafología y Química Agrícola, Facultad de Biología, Universidad de Santiago de Compostela
2
Laboratory of Soil Science and Geology, Wageningen University, Wageningen, The Netherlands
Within the framework of EU-COST action 622 twenty soils developed on volcanic materials
from Italy (EUROl to 04), Azores Islands (EUR05, 06), Iceland (EUR07 to 09), Canary
Islands (EUR10 to 12), Greece (EUR13 to 15), France (EUR16, 17) and Hungary (EUR18 to
20) were described, sampled and analyzed for a large number of soil properties by different
research groups. This database provides a good opportunity for the application of multivariate
statistical methods, although the number of samples is moderate (94). We developed two
different approaches for the explanation of the variance structure of the European volcanic
soils: 1) an exploratory analysis using principal component analysis (PCA), and 2) a
confirmatory analysis based on PCA separation and knowledge on andic and vitric horizons
using discriminant analysis.
Table 1 Soil properties of European volcanic soils used for
The soil properties used for the
analyses are listed in Table 1, and
relate to aspects like granulometry,
soil reaction, organic matter, and
parameters characterizing the
reactive components. As mentioned
above, in a first step we performed
an exploratory PCA analysis. Six
axes explained 84% of the total
variance and the first two a 60%.
From the projection of these first
two axes as well as from the
components matrix it was clear that
the first factor is dominated by
properties related to reactive
components at one side and soil
reaction, exchange complex (SB,
Bsat) and the composition of
allophane (Alp-Alo/Sio) at the other
side; while the variation in the
second component is dominated by
the organic matter and its influence
in other soil properties (C, Alp, Fep, Alp/Alo, CEC). The clay content groups with the last set
of properties, perhaps due to its effect on soil CEC.
The plot of the samples in the projection of these two first axes (Figure 1) shows a
separation of andic and vitric horizons. Non-andic/vitric horizons plot in the left side while
vitric and silandic horizons spread to the right side, but remain separated. The maximum of
reactive components is represented by all horizons of soil EUR06 (Azores). Organic and
aluandic horizons appear above the silandic horizons. Their small number does not result in a
clearly separated group.In addition,the Alu-andic horizons are very rich in organic matter.
Granulometry clay
Soil reaction pH in water (pHw), pH in KCl (pHk), pHk-pHw
Organic matter total C content
Exchange complex CEC, Base cations (SB), base saturation (BSat)
Reactive components Total A1 and Fe (Alt, Fet), extracted in NaOH (Aln, Sin), extracted in acid aommonium oxalte (Alo, Feo, Sio), extracted in CuCF (Alcu), extracted in Na-pyrophosphate (Alp, Fep), extracted in dithionite-citrate (Fed).
pHNaF (pHf), P rentention
Calculated variables: Alo+l/2Feo, Alo/Alt; Alp/Alt, Alp/Alo; (Alo-Alp)/Sio, Fed/Fet, Feo/Fed, Fep/Feo; allophane content
10