engleski

Evaluation of spatial and temporal variation in water contamination along Croatian highways by multivariate exploratory analysis

In the research in this paper we investigate spatial and temporal variations in the composition of wastewater near Croatian highways in three climatic regions (continental, Mediterranean, highland) during three seasons (autumn, winter and spring). In our paper the spatial division of the investigated areas that pertain to the three afore-mentioned climatic regions was obtained using the method of hierarchical clustering of monitored locations. 1533 samples from 14 locations along Croatian highways were collected and analysed by methods of multivariate exploratory analysis. By methods of principal components, factor analysis and hierarchical clustering of variables we grouped the variables into factors. Whereas 60% of variation in the data was explained by three principal components, six principal components accounted for 88% of data variation. The key section of our research was conducted by the decision trees method. For the purpose of analysis we classified 1533 samples into three classes representing climatic regions separately for each season and obtained the accuracy of 76-90% on test samples. Finally, using decision trees we identified the most important variables that differentiate climatic regions by the level of contamination of water along highways in different seasons.

highway; water contamination; hierarchical clustering; principal components; factor analysis; decision tree

nije evidentirano