HR: 16:00h
AN: H54B-01 INVITED     [Abstracts]
TI: Challenges in Soft Computing: Case Study with Louisville MSD CSO Modeling
AU: * Ormsbee, L
EM: lormsbee@engr.uky.edu
AU: Tufail, M
EM: mtufail@engr.uky.edu
AB: The principal constituents of soft computing include fuzzy logic, neural computing, evolutionary computation, machine learning, and probabilistic reasoning. There are numerous applications of these constituents (both individually and combination of two or more) in the area of water resources and environmental systems. These range from development of data driven models to optimal control strategies to assist in more informed and intelligent decision making process. Availability of data is critical to such applications and having scarce data may lead to models that do not represent the response function over the entire domain. At the same time, too much data has a tendency to lead to over-constraining of the problem. This paper will describe the application of a subset of these soft computing techniques (neural computing and genetic algorithms) to the Beargrass Creek watershed in Louisville, Kentucky. The application include development of inductive models as substitutes for more complex process-based models to predict water quality of key constituents (such as dissolved oxygen) and use them in an optimization framework for optimal load reductions. Such a process will facilitate the development of total maximum daily loads for the impaired water bodies in the watershed. Some of the challenges faced in this application include 1) uncertainty in data sets, 2) model application, and 3) development of cause-and-effect relationships between water quality constituents and watershed parameters through use of inductive models. The paper will discuss these challenges and how they affect the desired goals of the project.
DE: 1847 Modeling
DE: 1871 Surface water quality
DE: 1879 Watershed
SC: Hydrology [H]
MN: Fall Meeting 2005