variavel0=<b>Renato Simoes Silva</b> - rssr@lncc.br Lab. Nacional de Computacao Cientifica
<b>Fernanda Maria Pereira Raupp</b> - fernanda@lncc.br Laboratorio Nacional de Computacao Cientifica
<b>Regina Celia Almeida</b> - rcca@lncc.br Laboratorio Nacional de Computacao Cientifica
<b>Abstract.</b> The advances in computer technologies have expanded the capability for numerical  simulation in science and engineering. This happen specially in environmental problems since their mathematical and computational complexity can grow very fast. Their solutions have required the development of appropriate mathematical models, numerical methodologies and algotithms to solve practical problems. A common environmental problem is the thermal pollution created by a thermal powel plant where cold water is pumped into the plant and returned at elevated temperature. This relatively hot water is detrimental to aquatic life in rivers, lakes or bays. There are two approaches to deal with thermal pollution problems. In the near-field approach the problem is concerned with a very limited zone, near the discharge point. The other one is the far-field approach in which the main emphasis is on the effects of a perturbation at a considerable distance doenstream. The purpose of this work is to provide a numerical methodology to solve thermal pollution problems efficiently in the far-field approximation. The Streamline Upwind Petrov-Galerkin Method provides the numerical stability of the finite element method. The solution of the algebraic linear system that arises from the finite element discretization is performed by a new method called Left Conjugade Gradient Method presented by Yuan, Golub and Pelmmons. Its performance is compared with the most traditional methods for this type of problems, the GMRES Method.
<b>Keywords.</b> thermal pollution, convection-diffusion, iterative methods, finite element.