Variability of estimated contamination extent depending on calculation methods

Abstract

<p>The aim of research was to analyse the changes of contamination extent (E, %) estimated using two methods of background calculation by eliminating anomalies outside the interval mean±2s or median±2MAD (s is standard deviation, MAD is median absolute deviation) and optional normalisation. Two methods were used for estimation of background values (B) and upper threshold values (T) of Ag, As, Ba, Co, Cr, Cu, Mn, Mo, Ni, Pb, Sn, U, V, Zn and selected major elements according to their contents in topsoil samples from football fields in less contaminated districts of Klaipėda. Then two different sets of B and T values were applied to the whole geochemical data set. After determination of two sets of concentration coefficients (K<sub>k</sub>), four sets of enrichment factors were calculated: two simple (EF<sub>1</sub>) (normalisation by Al) and two complex (EF<sub>3</sub>) (normalisation by Al, K, Ti). </p><p>Estimation of E using T values resulted in lower percentage compared to percentage of sites where K<sub>k</sub>>1, EF<sub>1</sub>>1 or EF<sub>3</sub>>1, because the latter 3 indices depend on B values. Since all T values obtained by median±2MAD method are much lower (by 6-37.1%) than by mean±2s method, respective E is much higher: for 11 pollutants it is higher by more than 5% (range 6.3-34.2%). Since the absolute difference between B values estimated by two methods is much lower (only for Sn and Mn exceeds 5%) than the difference between T values, the influence of different B values on estimated E is much lower. Higher than 5% absolute difference between two sets of E estimated according to K<sub>k</sub>>1 is observed for 5 pollutants, according to EF<sub>3</sub>>1 for 4 and according to EF<sub>1</sub>>1 only for 3 pollutants. So E estimated according to EF<sub>1</sub>>1 is least of all influenced by the method of determination of B values, besides, it usually gives the highest E.</p>