Teams of aerial robots can perform sensing and surveillance missions faster and with more reliability than single-robot systems. The convergence of wireless networked communication, embedded systems, miniaturized sensors, and artificial intelligence has made it practical to design and deploy teams of aerial robots in complex environments such as severe local storms. This presentation describes recent advances toward the design, implementation, and deployment of teams of fixed-wing unmanned aircraft for various autonomous information gathering missions. An autonomous decision-making architecture is presented that exploits access to the Internet and dispersed computing resources in the decision-making loop. Aspects of the architecture are presented in the context of applications that include cooperative tracking of moving ground targets using radio emissions and computer vision, planning through complex wind fields, and studying tornado formation in supercell thunderstorms.