Owing to the increasing frequency of rainstorm disasters, strengthening the link between meteorological warning and emergency response that is early warning response has become crucial. However, early warning response usually fail, the underlying reasons behind this phenomenon and their improvement measures need to be urgently explored. This study uses the principal–agent theory to establish an incentive-constraining model based on hierarchical government to explore this problem. Considering the interest game between the superior and local governments, we mainly discuss how to design the optimal incentive mechanism under the condition of information asymmetry to enhance the local government's enthusiasm for providing an early warning response to rainstorm, thereby maximizing the benefits for both parties in the model. Furthermore, based on actual conditions, we use numerical simulation to present the model results more clearly. The results indicate that incentive and supervision mechanisms are mutually substitutable. Strong early warning capability and active public cooperation can improve the local government's enthusiasm for providing an early warning response. Excessive costs lead the local government to adopt a negative attitude toward warning, regardless of supervision is introduced. Only when the local government exhibits strong risk aversion and its investment is affected by numerous random factors, introducing supervision is effective. Meanwhile, too many external factors, excessive or insufficient supervision's intensity also affect the optimization effect of supervision on incentive contract. To maximize the effectiveness of supervision, the superior government should improve the supervision team's capability, and maintain the supervision's intensity within a reasonable range.