Application scheduling in Grid computing requires information about tasks in each application, as well as a description of the available resources that the scheduler needs, to find an efficient assignment of tasks to resources. In this context, information about applications is provided by the application model. Most current models are restricted to only a few features, but in some cases, such as simulations, which are an important example of Grid applications, more detailed information is available. In this paper, we propose an enhanced application model based on the concept of task refinements, which provide the scheduler with fine-grained information and allow it to determine more efficient schedules than traditional approaches. We backup and show the feasibility of our model by means of experimental evaluations.