Emergency responders require accurate geographic information to locate and assess incidents reported on social media. Unfortunately, however, social media often lacks geographic metadata or includes geotags that inaccurately reflect the locations of reported events. Toponyms, place names in social media content, provide another source of geographic information yet toponym usage in emergencies remains poorly understood.To address this gap, this study examines toponym usage across 22,343 tweets posted during a severe storm in the Northeastern United States. Tweets were qualitatively coded for nineteen types of storm-related information as well as toponym usage, location reference, and granularity. Findings show that users i) include more geographic information, including toponyms and geotags, in tweets about emergencies than about other topics; ii) report geographically distributed events but tend to include toponyms rather than geotags when reporting local information; and iii) include mostly municipal and regional toponyms in tweets about emergencies but tend to include hyperlocal toponyms and additional municipal or regional markers in reports of infrastructure damage and service disruption.Together these findings offer implications for social sensing applications in emergency response and management. These include more opportunities for course-grained damage assessment than fine-grained situational awareness, and the need for internal and external sources of geographic information to distinguish information about local and remote events. Lastly, findings suggest opportunities for passive and active sensing approaches responsive to the recipient design of social media: users provide additional geographic context that help others distinguish local and remote risks within noisy information spaces that emerge during geographically distributed crises.