The topology of vector fields offers a well known way to show a "condensed" view of the stream line behavior of a vector field. The global structure of a field can be shown without time-consuming user interaction. With regard to large data visualization, one encounters a major drawback: the necessity to analyze a whole data set, even when interested in only a small region. We show that one can localize the topology concept by including the boundary in the topology analysis. The idea is demonstrated for a turbulent swirling jet simulation example. Our concept works for all planar, piecewise analytic vector fields on bounded domains. ÁÒØÖÓ Ù Ø ÓÒ Vector fields are a major "data type" in scientific visualization. In fluid mechanics, velocity and vorticity are given as vector fields. This holds also for pressure or density gradient fields. Electromagnetics is another large application area with vector fields describing electric and magnetic forces. In solid mechanics, displacements are typical vector fields under consideration. Science and engineering study vector fields in different contexts. Measurements and simulations result in large data sets with vector data that must be visualized. Besides interactive and texture-based methods, topological methods have been studied by the visualization community, see Helman and Hesselink, 1990, Globus et al., 1991 for example. In most cases, the scientist or engineer is in-½ ¾ terested in integral curves instead of the vector field itself. Since the behavior of curves differs, a natural approach is to study classes of equivalent curves. This approach reduces the information concerning all curves to the information about the structural changes of curves. Vector field topology is one answer to this question. First, one detects all stationary curves, i. e., the critical points (or zeros) in the vector field. Then, one finds all integral curves where the behavior is different between the neighboring curves. These "separatrices" are then visualized together with the critical points, providing a detailed description of the behavior of the integral curves. A major drawback of topology-based visualization is the fact that one must analyze the whole data set. In many situations, a scientist or engineer would like to understand the behavior of curves in a limited area only. Due to the global nature of topology, one analyzes, up to now, the whole data set to find all separatrices in this area. In this paper, we show that this is not necessary. By a strict topological analysis of the boundary (of the local region of interest), we find all structural changes of the integral curves in any bounded region without touching data outside the region. We start by providing a rigorous mathematical treatment of these concepts. We continue with algorithmic details and conclude with theoretical and practical examples.