IN21A-3697:
Spatial Pattern of Great Lakes Estuary Processes from Water Quality Sensing and Geostatistical Methods
Tuesday, 16 December 2014
Wenzhao Xu1, Barbara S Minsker1, Barbara Bailey2 and Paris Collingsworth3, (1)University of Illinois at Urbana Champaign, Civil and Environmental Engineering, Urbana, IL, United States, (2)San Diego State University, Department of Mathematics and Statistics, San Diego, CA, United States, (3)Environmental Protection Agency Chicago, Chicago, IL, United States
Abstract:
Mixing of river and lake water can alter water temperature, conductivity, and other properties that influence ecological processes in freshwater estuaries of the Great Lakes. This study uses geostatistical methods to rapidly visualize and understand water quality sampling results and enable adaptive sampling to remove anomalies and explore interesting phenomena in more detail. Triaxus, a towed undulating sensor package, was used for collecting various physical and biological water qualities in three estuary areas of Lake Michigan in Summer 2011. Based on the particular sampling pattern, data quality assurance and quality control (QA/QC) processes, including sensor synchronization, upcast and downcast separation, and spatial outlier removal are first applied. An automated kriging interpolation approach that considers trend and anisotropy is then proposed to estimate data on a gridded map for direct visualization. Other methods are explored with the data to gain more insights on water quality processes. Local G statistics serve as a supplementary tool to direct visualization. The method identifies statistically high value zones (hot spots) and low value zones (cold spots) in water chemistry across the estuaries, including locations of water sources and intrusions. In addition, chlorophyll concentration distributions are different among sites. To further understand the interactions and differences between river and lake water, K-means clustering algorithm is used to spatially cluster the water based on temperature and specific conductivity. Statistical analysis indicates that clusters with significant river water can be identified from higher turbidity, specific conductivity, and chlorophyll concentrations. Different ratios between zooplankton biomass and density indicate different zooplankton structure across clusters. All of these methods can contribute to improved near real-time analysis of future sampling activity.
