The use of workflow techniques in scientific computing is widely adopted in the execution of experiments and building in silico models. By analysing some challenges faced by a scientific application in the geosciences domain, we noticed that workflows could be used to represent the geological models created using the application so as to ease the development of features to meet those challenges. Most works and tools on the scientific workflows domain, however, are designed for use in distributed computing contexts like web services and grid computing, which makes them unsuitable for integration or use within simpler scientific applications. In this dissertation, we discuss how to make viable the composition and representation of workflows within an existing scientific application. We describe a conceptual architecture of a workflow engine designed to be used within a stand-alone application. We also describe an implementation model of this architecture in a C plus plus application using Petri nets to model a workflow and C plus plus functions to represent tasks. As proof of concept, we implement this workflow model in an existing application and studied its impact on the application.