A0462
Title: Geographically weighted principal components analysis and variography for environmental variables
Authors: Monica Palma - University of Salento (Italy) [presenting]
Sabrina Maggio - University of Salento (Italy)
Giuseppina Giungato - University of Salento (Italy)
Abstract: Several statistical techniques of multivariate analysis, such as principal component analysis (PCA), allow the researchers to build very simple way composite indicators measuring the phenomenon under study. In social and environmental sciences, it is very common to synthesize the variables of interest in a unique indicator suitable to describe the variables at hand and which could be used by policymakers as decision support. In the presence of a multivariate geo-referenced data set, the classical multivariate techniques are not at all adequate to define composite indicators. In this case, the geographically weighted PCA (GWPCA) is a valid approach to constructing spatial composite indicators, taking into account the multivariate spatial dependence of the study variables. Recently, a new approach to choosing one of the GWPCAs parameters, i.e., the bandwidth of the kernel weighting function, has been proposed in a recent study and applied in a socio-economic context. The novel approach for GWPCA is used to construct a spatial composite indicator of urban air quality over a risk area, and the appropriateness of the found indicator in estimating the environmental quality at un-sampled locations is also discussed.--Funding--This paper has been supported by Consorzio Universitario Interprovinciale Salentino (CUIS) of the Province of Lecce (Italy).
