Ground Water
| Ideas |
When developing models, one of the most important questions is 'How should our resources be spent if the goal is a model that accurately represents the real system?' In groundwater hydrology and many other fields, scientists and enginieers deal with problems for which the quantities needed to construct the models can not generally be measured accurately. Under such circumstances, it is important to design a model that can be used to test hypotheses about the system against available observations of model-calculated quantities, such as hydraulic head and flows in ground-water flow models. Such testing requires many model simulations, and typically there is substantial competition between this goal and the goal of representing additional system processes.
While there is no single answer to this conflict that applies to all situations, a rule of thumb that often is helpful is to limit forward model execution times to about 30 minutes or less. Given present computer capabilities, this is sufficient for the simulation of many ground-water flow problems, but is a significant problem related to other situations, such as conservative and reactive transport and multiphase flow. Considering the rapid changes in computing capabilities, this is expected to become much less of a problem in the next 10 years. Also, the calculation of sensitivities is highly parallelizable. Even just parallelizing the calculation of sensitivities, which is relatively easy, allows convenient analysis with execution times of an hour or more.
Computer execution time issues are discussed in more detail in the Methods and Guidelines Report.
| Accessibility|FOIA|Privacy|Policies and Notices | |
  |
|
