- 5:20 PM
422f

Water Height Prediction in Mobile Bay Using Wavelet-Based Multi-Scale Model

Katherine Thorn1, Kyeong Park2, Nicholas D. Sylvester1, and Manish Misra1. (1) Chemical Engineering, University of South Alabama, Engineering Lab Building EGLB 248, Mobile, AL 36688, (2) Marine Sciences, University of South Alabama, Mobile, AL 36688

Mathematical modeling of the water height in ecological systems has recently become important due to increased abnormality in earth's weather pattern, which has caused numerous hurricanes and enhanced rainfall events. It is desirable to develop a mathematical model that can predict the water height by accounting for variations in the local weather conditions. For most water bodies, water height is dependent on the type of the water body, and on the environmental parameters such as tides, wind speed and direction, rainfall, freshwater discharge, etc. These causing parameters affect the water level in varying degrees. A suitable formulation for predicting water height should judiciously incorporate these parameters. As such, the problem of developing a mathematical model for water height prediction is non-trivial. An optimal approach seems to utilize a modeling technique which can capture the effects of various causing parameters in a multi-scale formulation, where contributions from parameters that affect the water height at different rates are incorporated at different time scales in the model.

Mobile Bay in southern Alabama is an estuary which is ecologically different from other areas that have been modeled. This study focuses on developing a multi-scale model for predicting water height in and around Mobile Bay. A multivariate array of time series data from relevant environmental parameters were input to a wavelet-based multi-scale model. The wavelet model was used to decompose each data stream into components at different time-scales. Wavelet components which provided insignificant or irrelevant information towards water height prediction were ignored. Use of wavelet ensured that process noise and disturbances, which can often cause a nuisance while developing ecological models, were removed. The wavelet decomposed data revealed the core trend of the environmental parameters and a multi-scale model was developed by judicious combination of wavelet decomposed data of environmental parameters. Tests with data, sampled hourly for seven months, demonstrated superior performance by the wavelet-based multi-scale model as compared with conventional time-series modeling approaches.