We present a novel technique for analyzing the geometry of streamlines representing large scale flow fields produced in scientific simulations. We introduce the box counting ratio, a metric related to the Kolmogorov capacity or box counting dimension, for quantifying geometric complexity of streamlines (or streamline segments). We utilize this metric to drive a visual analytic framework for extracting, organizing and representing features of varying sizes from large number of streamlines. This framework allows the user to easily visualize and interact with the features otherwise hidden in large data. We present case studies using combustion and climate simulation datasets.