Application of Chemometric Analysis to the Study of Snow at the Sudety Mountains, Poland

Snow samples were collected during winter 2011/2012 in three posts in the Western Sudety Mountains (Poland) in 3 consecutive phases of snow cover development, i.e. stabilisation (Feb 1st), growth (Mar 15th) and its ablation (Mar 27th). To maintain a fixed number of samples, each snow profile has been divided into six layers, but hydrochemical indications were made for each 10 cm section of core. The complete data set was subjected in the first run of chemometric data interpretation to Cluster Analysis as well as Principal Components Analysis. Further, Self-Organizing Maps, type of neutral network described by Kohonen were used for visualization and interpretation of large high-dimensional data sets. For each site the hierarchical Ward’s method of linkage, squared Euclidean distance as similarity measure, standardized raw data, cluster significance test according to Sneath’s criterion clustering of the chemical variables was done. Afterwards this grouping of the chemical variables was confirmed by the results from Principal Components Analysis. The major conclusion is that the whole system of three sampling sites four patterns of variable groupings are observed: the first one is related to the mineral salt impact; the second one - with the impact of secondary emissions and organic pollutants; next one - with dissolved matter effect and the last one - with oxidative influence, again with relation to anthropogenic activities like smog, coal burning, traffic etc. It might be also concluded that specificity of the samples is determined by the factors responsible for the data set structure and not by particular individual or time factors.