HR: 1340h
AN: H13A-0976    [Abstracts]
TI: Integrating Sensor Data and Informatics to Improve Understanding of Hypoxia in the WATERS Network Testbed at Corpus Christi Bay, Texas
AU: Coopersmith, E
EM: ecooper2@uiuc.edu
AF: University of Illinois at Urbana-Champaign, Civil and Environmental Engineering 4230 Newmark Lab MC-250, 205 N. Mathews Ave., Urbana, IL 61801, United States
AU: * Kulis, P
EM: paulakulis@mail.utexas.edu
AF: University of Texas at Austin, CRWR Pickle Research Campus Bldg 119, MC R8000, Austin, TX 78712, United States
AU: Brouwer, A
EM: brouwe70@potsdam.edu
AF: State University of New York at Potsdam, 517 Barrington Dr., Potsdam, NY 13676, United States
AU: Montagna, P
EM: paul.montagna@tamucc.edu
AF: Texas A&M University at Corpus Christi, Harte Research Institute room 210A 6300 Ocean, Unit 5869, Corpus Christi, TX 78412, United States
AU: Hodges, B R
EM: hodges@mail.utexas.edu
AF: University of Texas at Austin, CRWR Pickle Research Campus Bldg 119, MC R8000, Austin, TX 78712, United States
AU: Minsker, B
EM: minsker@uiuc.edu
AF: University of Illinois at Urbana-Champaign, Civil and Environmental Engineering 3230d Newmark Lab, MC-250 205 N. Mathews Ave., Urbana, IL 61801, United States
AU: Maidment, D
EM: maidment@mail.utexas.edu
AF: University of Texas at Austin, CRWR Pickle Research Campus Bldg 119, MC R8000, Austin, TX 78712, United States
AB: The goal of the WATERS Network Testbed in Corpus Christi Bay (Texas) is to better understand hypoxia by creating a prototype Environmental Information System (EIS) that links field data collection, real-time modeling techniques, and cyberinfrastructure. In this paper, we explore the connection between the bay's bottom-water hypoxia and wind mixing by integrating several field data sets within a machine-learning model and exploring the mechanisms leading to the model results using an independent data set. K-nearest neighbor machine learning models applied to several long-term data sets indicate that wind velocities are instrumental in forecasting hypoxic events. Additionally, statistical analysis suggests that the impacts of wind vary spatially throughout the bay. Forecasting algorithms can be employed to predict not only the expected value of dissolved oxygen levels throughout the bay, but also the probability of observing hypolimnetic hypoxia. Prior values of dissolved oxygen, salinity, wind direction, wind velocity, and water temperature have been shown to play a meaningful role in influencing the DO value twenty-four hours hence. Visualizing spatial maps of expected means and variances not only illustrate potentially hypoxia regions, but areas where future sampling would be most beneficial as well. We use a short-term field data set to explore the possible mechanisms controlling the observed statistical trends in long-term data sets. Field data taken from July 2006 document a specific hypoxic episode that follows a high wind event. Analyses of temporal changes in the vertical water column support the suspected connections between wind, salinity, and hypoxia, and suggest some possible mechanisms for this connection. It is suspected that wind controls the sinking of heavy, saline water into the bottom of Corpus Christi Bay from Laguna Madre, a nearby shallower bay. This isolation of dense water from surface oxygen replenishment may be critical in hypoxia development. Field data also suggest that subsequent water column mixing (following hypoxia formation) is controlled by the wind. The link between sophisticated statistical models and mechanistic analysis is used to support both analysis methods and independent hypotheses, and also to provide new insights. Many of the analyses conducted on the short-term data set are influenced by the results from our statistical models, and results from the mechanistic analysis has influenced some statistical analysis.
DE: 1816 Estimation and forecasting
DE: 1847 Modeling
DE: 1848 Monitoring networks
SC: Hydrology [H]
MN: 2007 Fall Meeting