Computational Hydrology

Scope:

Research in computational hydrology aims at improving the mathematical and numerical methods used to predict the behavior of complex hydrologic systems.  Simulating subsurface processes requires that the governing equations be discretized in space and time using numerical schemes that are accurate, robust, and efficient.  Algorithms have to be developed and adapted to the characteristics of the equations, the requirements of specific applications, and the available computational resources.  Advanced simulation techniques are essential for the study of coupled hydrological systems on multiple scales, and to enable application of numerical modeling to challenging prediction and optimization problems in water resources, energy recovery, and waste isolation projects.

ESD Activities:

The iTOUGH2 code developed in ESD is the inverse modeling framework for TOUGH2 capable of carrying out inverse modeling, optimization, and uncertainty analyses.  Because iTOUGH2 must make repeated calls to TOUGH2, its computational demands are high.  In addition, the growing use of coupled models (e.g., TOUGHREACT, TOUGH-FLAC) is placing further demands on computational efficiency.  With the growth in demand for fast simulations of complex processes and for inverse modeling, the need for efficient algorithms has never been greater.

The TOUGH code developed in ESD for the simulation of nonisothermal, multiphase, multicomponent flow and transport in fractured porous media is currently being re-engineered to extend, couple, streamline, and unify processes and capabilities which are presently implemented in multiple modules.  The objective of this effort is to provide a solid, transparent, and uniform computing platform with a flexible data structure, modular architecture, and coherent implementation of features as the basis for future code developments.

Techniques for the joint inversion of geophysical and hydrological data are being investigated to improve characterization of highly heterogeneous systems.  Automatic differencing and streamline approaches are being considered as alternatives to the conventional methods for calculating sensitivity coefficients.  Finally, parallelization and grid computing techniques are being examined, including the development of a massively parallel version of TOUGH2 called TOUGH2-MP, capable of solving problems with more than 1 million gridblocks.

Contact:

Stefan Finsterle
ph: 510.486.5205
email: safinsterle@lbl.gov