Public communication is critical for responding to disasters. However, most research on public communication is largely focused on its informational support function, overlooking the emotional support that could equally offer. This study takes the lead to investigate their separate impacts. In particular, the variable public engagement, which is a function of the number of Shares, Likes, and Comments in a particular post, is introduced to benchmark the effect of public communication. Besides, considering the evolving nature of the crisis, their dynamic impacts across different COVID-19 pandemic stages are examined. Data from Dec 2019 to Jul 2020 were collected from 17 provincial government-owned social media (Weibo) accounts across COVID-19 in China with a Natural Language Processing-based method to compute the strengths of informational support and emotional support strength. An econometric model is then proposed to explore the impacts of two supports. The findings are twofold: the impact of emotional support on public engagement is empirically confirmed in the study, which is not in lockstep with the informational support; and their impacts on public communication are dynamic rather than static across stages throughout the crisis. We highlighted the importance of emotional support in public engagement by deriving its impact separately from informational support. The findings suggest incorporating both social supports to create stronger public communication tactics during crises.