Quantifying Linkages between Biogeochemical Processes in a Contaminated Aquifer-Wetland System Using Multivariate Statistics and HP1

Fate and transport of contaminants in saturated and unsaturated zones in the subsurface is controlled by complex biogeochemical processes such as precipitation, sorption-desorption, ion-exchange, redox, etc. In dynamic systems such as wetlands and anaerobic aquifers, these processes are coupled and can interact non-linearly with each other. Variability in measured hydrological, geochemical and microbiological parameters thus corresponds to multiple processes simultaneously. To infer the contributing processes, it is important to eliminate correlations and to identify inter-linkages between factors. The objective of this study is to develop quantitative relationships between hydrological (initial and boundary conditions, hydraulic conductivity ratio, and soil layering), geochemical (mineralogy, surface area, redox potential, and organic matter) and microbiological factors (MPN) that alter the biogeochemical processes at the column scale. Data used in this study were collected from controlled flow experiments in: i) two homogeneous soil columns, ii) a layered soil column, iii) a soil column with embedded clay lenses, and iv) a soil column with embedded clay lenses and one central macropore. The soil columns represent increasing level of soil structural heterogeneity to better mimic the Norman Landfill research site. The Norman Landfill is a closed municipal facility with prevalent organic contamination. The sources of variation in the dataset were explored using multivariate statistical techniques and dominant biogeochemical processes were obtained using principal component analysis (PCA). Furthermore, artificial neural networks (ANN) coupled with HP1 was used to develop mathematical rules identifying different combinations of factors that trigger, sustain, accelerate/decelerate, or discontinue the biogeochemical processes. Experimental observations show that infiltrating water triggers biogeochemical processes in all soil columns. Similarly, slow release of water from low permeability clay lenses sustain biogeochemical cycling for a longer period of time than in homogeneous soil columns. Preliminary results indicate: i) certain variables (anion, cation concentrations, etc.) do not follow normal or lognormal distributions even at the column scale, ii) strong correlations exist between parameters related to redox geochemistry (pH with S2- concentrations), and iii) PCA can identify dominant processes (e.g. iron and sulfate reduction) occurring in the system by grouping together causative variables (e.g. dominant TEAPs).