At a global level, climate change is expected to result in more frequent and higher-intensity weather events, with impacts ranging from inconvenient to catastrophic. The potential for disasters to act as "focusing events" for policy change, including adaptation to climate change risk, is well known. Moreover, local action is an important element of climate change adaptation and related risk management efforts. As such, there is a good reason to expect local communities to mobilize in response to disaster events, both with immediate response and recovery-focused activities, as well as longer-term preparedness and adaptation-focused public policy changes. However, scholars also note that the experience of disaster does not always yield policy change; indeed, disasters can also result in policy inertia and failure, perhaps as often or more often than major policy change. This study poses two key research questions. First, we ask to what degree policy change occurs in communities impacted by an extreme weather event. Second, we seek to understand the conditions that lead to adaptation-oriented policy adoption in response to an extreme weather event. Our results suggest two main recipes for future-oriented policy adoption in the wake of an extreme weather event. For both recipes, a high-impact event is a necessary condition for future-oriented policy adoption. In the first recipe for change, policy adoption occurs in Democratic communities with highly focused media attention. The second, less expected recipe for change involves Republican communities that have experienced other uncommon weather events in the recent past. We use a comparative case approach with 15 cases and fuzzy set qualitative comparative analysis methods. Our approach adds to the existing literature on policy change and local adaptation by selecting a mid-N range of cases where extreme weather events have the potential to act as focusing events, thereby sidestepping selection on the dependent variable. Our approach also takes advantage of a novel method for measuring attention, the latent Dirichlet allocation approach.