This research places disaster management into a geographic context in order to understand how people communicate during disasters and how to increase local community resilience to foster sustainability and lessen vulnerability. Analyzing communication that occurred over Twitter networks during the 2014 San Diego wildfires and the 2015 Nepal earthquake enabled temporal trends, spatiotemporal patterns, social network structures, socially disconnected communities, and hyperlocal relationships to be identified and measured. The temporal trends examined daily communication by day and by hour to see how well Twitter activity correlated to real world events. In both case studies, there were strong relationships between Twitter activity and real world events; their major peaks and falls reflected occurrences recorded in the After Action Reports provided by the County of San Diego and the Government of Nepal. The spatiotemporal analysis showed the spatial distribution of tweets, and measured changes in communication and movement patterns. The results from the two case studies were similar in that their data produced major hotspots not only in highly populated regions, but also areas who were either directly affected or nearby affected areas. The differential maps showed that when each disaster event occurred, there would be a spike in Twitter activity in or around the disaster. Methods in social network analysis were performed on the datasets to analyze the structure of the online communities from each case study and to identify their socially disconnected communities. Both San Diego and Nepal's online Twitter networks had similar structures and measurements of modularity, the measurement that forms online communities and indicates the connectedness of a network. The similar measurements are a sign that there are dense connections between people, but fewer connections between communities. This information suggests that methods to improve communication between communities should be established. Recommendations are to create a uniform way for people to communicate during a disaster, such as designating specific hashtags or have influential people act as online sirens to carry any disaster-related information further into the networks. In addition to similar structures and measurements of modularity, both case studies' socially disconnected communities expressed similar characteristics in how they were identified. Since there are similar ways to identify socially disconnected communities in these case studies, there is potential in providing a simple way for response agencies to identify these communities in the future. Once the communities can be identified, outreach programs can be expanded to ensure all people can gain situational awareness and real-time updates. The hyperlocal relationships between each case studies' online and offline communities were defined by separately mapping communities to see the users' geographic proximities to each other. The hyperlocal relationships on a city-level were strong, but going closer into a neighborhood-level shows weaker relationships. Trajectories of the users in the datasets were also mapped in order to better show human behavior and movement during disaster events from a micro-level analysis. The information provided by examining people's trajectories may help response agencies in future decision making when establishing shelters and distributing aid, and understanding people's response times to evacuation notices and why they may wait before evacuating.