The Gulf of Mexico (GOM) is of primary importance to both the United States (US) and Mexico. More then half of the waterborne US tonnage transits through its waterways, 93% of the US offshore oil production and about 98% percent of the gas production come from its waters. The population along its shorelines has been steadily increasing, totaling 46.7 million in 1999 for the US portion of the GOM, while the population along its tributaries and the Mississippi river in particular has been increasing as well. Anthropogenic impacts are an associated and growing concern as for example the GOM has the largest zone of human caused coastal hypoxia in the Western Hemisphere. To preserve the health of the GOM while accommodating growing human activities will require a dedicated and informed multinational stewardship. Given the size and the intricacies of the GOM ecosystem an essential component of a good decision making process will be the availability of extensive and consistent coastal, ocean and riverine observations and models. Such systems are being steadily developed and improved; the network of stations managed by the Texas A&M University-Corpus Christi Division of Nearshore Research (DNR) is one of the largest and most sophisticated. The presentation will describe the DNR network, its unique software and procedures and a set of growing Artificial Intelligence (AI) based models that take advantage of the large amount of real-time information to compute short term operational forecasts.

At present DNR manages 53 active monitoring platforms along the Texas coast including 29 platforms of the Texas Coastal Ocean Observation Network (TCOON) and 7 water level monitoring platforms for the National Ocean Service. DNR is also assisting a group of Mexican scientists from the Instituto Politecnico Nacional (IPN) and the Universidad Nacional Autonoma de Mexico (UNAM) to develop a water level monitoring network similar to TCOON along the Mexican Gulf coast called Red de Observaciones y Predictiones de Variables de Oceanicas (ROPVO). The overall DNR network provides real-time or near real-time coastal measurements such as water levels, wind speeds and wind directions, barometric pressures as well as other variables such as dissolved oxygen, salinity, water currents and wave climates depending on the station. The primary use of the data has been to establish tidal datums, but increasingly the network has provided data for many other uses including the commercial shipping industry, recreational boaters, sailors and windsurfers, the shrimp and fishing industry, marine construction, and decision-makers responsible for marine safety and emergency evacuation in the event of an approaching hurricane.

A distinctive feature of the DNR network is its unique data management software which provides data in real-time or near real-time to its sponsors and the community through the World Wide Web and through automated phone services. The software and procedures were developed on the principle that all user interaction with the data management system takes place via web-based interfaces. Such interactions include for example site visit and maintenance reports and chain-of-custody records. Sponsors, scientists and other potential users can access all DNR data in a variety of graphic and text-based formats from http://lighthouse.tamucc.edu/. The DNR software and hardware were built and combined with a primary goal of reliability. The data are housed in ordinary PC-base computers rather than sophisticated proprietary systems such that parts or whole systems can be replaced quickly if needed. The software is based on open source technologies such as Linux and Perl such that DNR is not subject to changes in proprietary systems and has the flexibility to replace software components as new technologies become available or as the needs of DNR’s sponsors evolve. DNR has started providing access to all its software under the General Public License and aims at co-developing future versions of the web-based software and procedures.

The growing availability of real-time data and extensive and consistent environmental time series allows for new modeling techniques. DNR has been taking advantage of these real-time data streams to develop AI based models. When the time series encompass most encountered conditions such models can be trained to quantify relationships between past observations and future outcomes. For cases such as point/local forecasts these models can have significant advantages over other models such as classic statistical models and Finite Element/Finite Differences models. These advantages include the ability to model non-linear relationships and the implicit inclusion of the boundary conditions and forcing functions. Also once the models are trained AI based predictions are computed virtually instantly. DNR has recently developed and implemented models to forecast in real-time water levels and storm surges for the coast of Texas. DNR has also implemented web based software to compute tidal constituents, harmonic forecasts and persistence forecasts for water levels as well as gap filling and spike detection algorithms. Models under development include more sophisticated ANN water level forecasts based on eta-12 and GFS atmospheric predictions, a model to predict sea breezes, a model to predict indicator bacteria counts in coastal recreational waters and models to forecast spring flows and water levels in a karst aquifer. General opportunities of combining the growing availability of real-time data with AI based models to make operational forecasts will be discussed.