(92h) Using Wavelets to Understand Dissolved Oxygen Variations in Mobile Bay

After obtaining the profile data from the physical parameters measured in Mobile Bay from July 1st to Sept 24th of 2004, it was determined that the variations of dissolved oxygen (DO) were too complex to analyze using time-scale methods. To understand the dynamics of dissolved oxygen variations it is necessary run wavelet analysis upon the original signal in order to obtain a comprehensive model for DO using a multivariate least partial squares technique. If this is done successfully, the model could be used to analyze DO variations based on the variations in the other measurable variables in a less complicated manner. Wavelets decompose data into frequency components, so that it is possible to analyze each frequency component with a resolution matched to its scale. However, a choice of a wavelet basis function must be decided by trial and error, this function affects the ?compactness? of wavelet decomposition. It is necessary to choose a basis function that applies proper de-noising and results in an applicable time-frequency analysis. Observations of the original signal conclude that the model is one dimensional, and the most efficient way of finding a proficient basis function is by using the MATLAB® wavelet toolbox menu and the Wavelet 1-D package. The Daubechies and the Symlets families of functions seem to provide the best results for a proper statistical analysis to take place. SAS institute's JMP® the statistical discovery software provides a powerful Multivariate Partial Least Squares method for determining a comprehensive model and testing its error against actual measured variable, in this case a specific wavelet transformed DO signal. Once a comprehensive model of DO is found and tested on multiple sets of relevant data signals to prove its reliability, expansive research is needed to discover methods of predicting DO dynamics.