1. Introduction application of Artificial Neural Networks (ANN) to forecast future water levels and improve on the presently inadequate harmonic forecasts. Accurate water level forecasts are of vital importance along the Texas coast as the waterways of the northern Gulf of Mexico play a critical economic role for a number of industries including shipping, oil and gas, tourism, and fisheries. The economic impact of these industries is not limited to the Gulf coast as for example more than 50% of the US tonnage reaching the US by waterways transits through the Gulf of Mexico. In the study area the Corpus Christi (CC) estuary (see Figure 1) is home to the fifth largest US port by tonnage, the Port of Corpus Christi. Astronomical forcing or tides are well tabulated; however water level changes along the Gulf coast are frequently dominated by meteorological factors whose impact is often greater than the tidal range itself (eg Cox et al. 2002a). The National Oceanic and Atmospheric Administration (NOAA) stated that “presently published predictions do not meet working standards” when assessing the performance of current predictions, a parameter closely related to water levels, for regular weather conditions in Aransas Pass and CC Bay, both part of the study area (NOAA 1991, NOAA 1994).

A comparison between harmonic forecasts and measured water levels is illustrated in Figure 2 for one of the TCOON stations of the study area, the Naval Air Station (NAS). As can be observed, the difference between harmonic and measured water levels is frequently larger than the tidal range itself. The water anomaly or difference between measured and harmonically forecasted water levels is presented in Figure 2b. The anomaly regularly shifts from positive to negative by about half a meter during frontal passages in winter spring and fall or Julian Day (JD) 0-120 and JD 280-365. Figure 2c displays the squared wind speed or wind pseudostress for the same period. The strong and shifting winds during frontal passages are well correlated with the water anomaly and wind is generally recognized as the main non-tidal forcing driving water level changes (Garvine 1985, NOAA 1991, NOAA 1994). The agreement between harmonic and measured water levels is typically much closer during the summer months with the exception of the passage of tropical storms. In 1998 this was illustrated by tropical storm Frances, which affected water levels in CC Bay from JD 245 to JD 260. Although the focus of this model is the forecast of water levels during frontal passages, data affected by tropical storms have not been removed and the performance of the model will be assessed during the passage of tropical storms.