Runoff nodes, located where stream segments merge to form a larger one, carry important structural geographic information. Extensive research has been done on runoff nodes, but there is still a lack of work on measuring the information content of runoff nodes across watersheds. This paper develops a novel index, called the "Feature Information Index of Runoff Nodes" (FIIRN), to assess the feature information content according to the Theory of Rough Sets (TRS). Firstly, this paper establishes the conception of, and the algorithm behind, FIIRN. Then, by exploring seven typical landforms of the Loess Plateau of China as case studies, values of FIIRN are calculated after extracting all the runoff nodes. Our experiment shows that FIIRN can effectively represent the feature information of watersheds, such as runoff order or evolutionary status. In summary, FIIRN will play a significant role in future watershed analysis.