Multivariate Statistics and Geostatistical Analyses of Metal Elements in Soil Multivariate Statistics and Geostatistical Analyses of Metal Elements in Soil of Waste Disposal Site

Abstract

In disposal site, municipal solid waste (MSW) decomposes and produces three components of solid; liquid (leachate) and landfill gas. Open dumping facilities release huge quantity of metal elements into the surrounding water bodies, underlying soil layer and atmosphere. The main focus of this study was to identify the correlations of metal elements in soil, possible sources of their generation and contamination, distribution of metal elements spatially as well as the level of contamination of soil of a disposal site. To these endeavors, sixty soil samples were collected at a depth of 0-30 cm from the existing ground surface from a selected waste disposal site at Rajbandh, Khulna, Bangladesh. These study periods covered both the dry season (March to May, 2016) and rainy seasons (June to August, 2016). In the laboratory, the relevant metal elements of Aluminium (Al), Arsenic (As), Barium (Ba), Calcium (Ca), Cadmium (Cd), Cobalt (Co), Chromium (Cr), Copper (Cu), Iron (Fe), Mercury (Hg), Potassium (K), Manganese (Mn), Sodium (Na), Nickel (Ni), Lead (Pb), Antimony (Sb), Scandium (Sc), Strontium (Sr), Titanium (Ti), Vanadium (V) and Zinc (Zn) were measured through the standard test methods. Furthermore, the spatial distribution of metal elements in soil is necessitated to explore their extents. Implementation of interpolation techniques can provide better prediction of the distribution of metal elements in soil with least prediction errors. To these attempts, conventional statistics such as K-S test, S-W test and normal QQ plot was performed using SPSS. Based on normal QQ plot, it was observed that almost all the metal elements were distributed normally except As in soil for both the dry and rainy seasons. The agglomerative hierarchical clustering (AHC) was performed using XLSTAT. Results of each other indicating these metal elements derived from the same generation sources. In addition, results of PCA and AHC depicted that almost all the metal elements in soil derived from anthropogenic/human activities; least number of metal elements from natural sources as well as from both the natural and anthropogenic sources. In this study, Geostatistical analysis such as inverse distance weighting (IDW), local polynomial interpolation (LPI), radial basis functions (RBF) and ordinary kriging (OK) was performed using ArcGIS. Furthermore, the cross validations of IDW with power 1 to 5, LPI with order 1 to 3, RBF with five kernal functions as well as OK with eleven distinct models were performed to select the best fitted model for further assessing the performance of these interpolation techniques. Based on least value of MAPE, IDW with power 1 to 5 and RBF with different kernel functions showed comparatively more accurate prediction than that of LPI and OK. Based on RI, IDW1 showed best performance followed by OK. Lastly, based on all indices (MAPE, RI, etc.), IDW1 showed the best technique for all metal elments. Moreover, produced prediction surface for all the interpolation techniques showed most of the contaminated hotspots was found near the centre of disposal site for all the metal elements. Semivariogram showed that almost all the metal elements were moderately correlatated spatially and least number of metal elements were strongly and weekly correlated. In this study, a network model was developed by ANN to predict and depict the validity of observed concentration of metal elements obtained from laboratory based MSE and R-value. It was found that the predicted values from ANN were almost same as obtained from laboratory. Finally, it can be concluded that this study will so guide for more efficient prediction of spatial distribution of metal elements with their possible generation sources, and to remedial measures regarding the contamination of soil of the waste disposal sites all over the world.