TY  - JOUR
AU  - Altdorff, Daniel
AU  - Dietrich, Peter
TI  - Combination of electromagnetic induction and gamma spectrometry using K-means clustering: A study for evaluation of site partitioning
JO  - Journal of plant nutrition and soil science
VL  - 175
IS  - 3
SN  - 1436-8730
CY  - Weinheim
PB  - Wiley-VCH
M1  - FZJ-2013-02638
SP  - 345 - 354
PY  - 2012
AB  - Today rapid survey methods of proximal soil sensing (PSS) provide an increasing number of different  and highly resolved data. These multidimensional data sets can lead to multilayered and  complex maps of parameters which are only indirectly related to soil properties and soil functions.  However, in applications usually just one clear elementary map is required. It is of increasing  importance to tackle this problem utilizing a cluster algorithm for the synthesis and reduction  of multidimensional input variables. The cluster algorithm provides a partitioning of the investigated  site whereby the units are characterized by the statistics of the PSS data. Therefore, the  question that arises is how suitable is the suggested partitioning in terms of the delineation of different  soil units. In this study, we investigate the suitability of cluster partitioning through a case  study at a medium-scale test site (≈ 50 000 m2). Two common PSS methods: electromagnetic  induction (EMI) and gamma spectrometry (GS) will be employed to create a data set for partitioning  by a K-means cluster. The result of the cluster analysis is a delineation of three different  parts. In contrast to previous studies, we evaluate the generated partitions by independent soil  properties such as grain size, horizon thickness, and color of stratified randomly taken soil samples.  The soil analyses show that one of three clusters significantly differs from the others in  terms of grain-size distribution and horizon thickness. The partitioning of the other two clusters  could not be confirmed by the considered soil parameters. Nevertheless, the case study demonstrates  the combination of different PSS data by K-means clustering as a potential approach for  site partitioning. An evaluation of the results of the cluster analysis through the collection and  analysis of soil samples is highly recommended.
LB  - PUB:(DE-HGF)16
UR  - <Go to ISI:>//WOS:000304598900003
DO  - DOI:10.1002/jpln.201100262
UR  - https://juser.fz-juelich.de/record/134439
ER  -