As large-scale environmental disasters become increasingly frequent and more severe globally, people and organizations that prepare for and respond to these crises need efficient and effective ways to integrate sound science into their decision making. Experience has shown that integrating nongovernmental scientific expertise into disaster decision making can improve the quality of the response, and is most effective if the integration occurs before, during, and after a crisis, not just during a crisis. However, collaboration between academic, government, and industry scientists, decision makers, and responders is frequently difficult because of cultural differences, misaligned incentives, time pressures, and legal constraints. Our study addressed this challenge by using the Deep Change Method, a design methodology developed by Stanford ChangeLabs, which combines human-centered design, systems analysis, and behavioral psychology. We investigated underlying needs and motivations of government agency staff and academic scientists, mapped the root causes underlying the relationship failures between these two communities based on their experiences, and identified leverage points for shifting deeply rooted perceptions that impede collaboration. We found that building trust and creating mutual value between multiple stakeholders before crises occur is likely to increase the effectiveness of problem solving. We propose a solution, the Science Action Network, which is designed to address barriers to scientific collaboration by providing new mechanisms to build and improve trust and communication between government administrators and scientists, industry representatives, and academic scientists. The Science Action Network has the potential to ensure cross-disaster preparedness and science-based decision making through novel partnerships and scientific coordination.