MULTIVARIATE STATISTICAL ANALYSIS OF VARIABILITY IN BIOGEOCHEMISTRY IN A CONTAMINATED AQUIFER-WETLAND SYSTEM

Steep biogeochemical gradients have been observed at hydrologic “mixing” interfaces in a wetland-aquifer system impacted by landfill leachate in Norman, Oklahoma. The wetland-aquifer system lies within the alluvial plain of the Canadian River and is characterized by silt layers interbedded with sandy layers, where exchange of groundwater and wetland water has been observed. Using cm-scale passive diffusion samplers (peepers), water samples were collected in a depth profile to span hydrologic interfaces between surface water and various sedimentary layers during three wet (spring) seasons (2003, 2004, and 2005) and a dry (fall) season (2005). Geochemical measurements were made on these samples including major electron acceptors (e.g., NO₃^- and SO₄^2-), electron donors (e.g., low molecular weight organic acids), other biogeochemical indicators (e.g., CH₄, NH₄⁺, Fe²⁺, S^2-, and DOC), mineral-water solubility indicators (e.g., Ca²⁺, Mg²⁺, K⁺) and solute transport indicators (e.g. Cl^-, leachate tracer). Relationships between geochemical parameters and sample locations were explored using multivariate statistics, including principal component factor analysis and cluster analysis. Relationships between biogeochemical parameters were interpreted to identify dominant processes such as terminal electron accepting processes (TEAPs) and mineral-water interactions. Relationships between sample locations were interpreted to identify dominant water masses and correlate to lithologic boundaries within the aquifer-wetland system. A Multivariate Analysis of Variance (MANOVA) test was performed to determine the main effects of factors such as recharge conditions and depth. Results show dominant TEAPs do not cluster into discrete redox zones because of heterogeneity in hydrologic interfaces and sediment geochemistry and, although TEAPs are temporally variable, do not appear to be directly related to recharge conditions. Our analyses suggest that graphical descriptions of the spatial and temporal variability in gradients are not always sufficient to illuminate the underlying controlling factors of biogeochemical heterogeneities. Multivariate statistical analysis is a powerful approach to explore variability in complex environmental datasets.