An interval fuzzy model for magnetic monitoring: estimation of a pollution index.

MARCOS A. E. CHAPARRO; MAURO A.E. CHAPARRO; ANA M. SINITO

Castle of Nove Hrady, South Bohemia, Czech Republic

Congreso; 12th Castle Meeting on Paleo, Rock and Environmental Magnetism 2010; 2010

Institute of Geophysics - Academy of Science of the Czech Republic

Pollution is a subject of current interest and there is a need for monitoring techniques developed by several fields of research. Although the man-made contribution of heavy metal and other pollutants can be studied by careful chemical methods (time-consuming, laborious and costly), magnetic monitoring constitutes an alternative tool for pollution studies. In our previous studies (Chaparro et al., 2008, Marié et al., 2010), multivariate statistical analyses were investigated for magnetic monitoring in soils and road-deposited sediments, revealing a link between magnetic and chemical variables. In this contribution, continuing the above-mentioned pollution studies, we report on a methodology in order to build an interval fuzzy model for the pollution index PLI (a composite index using Cr, Ni, Cu, Zn and Pb) in two dataset. In general, modelling based on fuzzy set theory is designed to mimic how the human brain tends to classify imprecise information or data. The interval fuzzy model seems to be an adequate mathematical tool for this nonlinear problematic. For this model, fuzzy c-means clustering is used to partition data, hence the membership functions and rules are built. In addition, interval arithmetic is used to get the fuzzy intervals. The studied sets are different examples of pollution by different anthropogenic sources, in two different study areas: (a) soil samples collected in Marambio station from Antarctica (n= 19); and (b) road-deposited sediments collected in the road Autovia 2 from Buenos Aires Province (n= 31). The datasets comprise magnetic and chemical variables, and for both cases, we selected the following relevant variables, magnetic concentrationdependant variables: mass-specific magnetic susceptibility (X) and anhysteretic remanent magnetisation (ARM); magnetic features-dependant variables: KARM/K-ratio and Hcr; and a chemical variable: PLI. The model output gives an estimation interval (its wide depends on the data density) for the measured values. The results show not only a good agreement between the estimation interval and data, but also provide valued information from the rules analysis that allows us to understand the magnetic behaviour of the studied variables under different conditions. From both studied cases, it is concluded that lower values of X, ARM (lower magnetic concentration), and KARM/K-ratio (coarser magnetic grain size) are associated with lower values of PLI (unpolluted samples). On the contrary, higher values of X, ARM (higher magnetic concentration), and KARM/K-ratio (finer magnetic grain size) are related to higher values of PLI (polluted samples)