This paper proposes a bi-level optimization model for resource allocation in disaster response, categorizing emergency resources into personnel and materials in consideration of the coupling relationship between them. The model addresses two levels of problems, which can be conceptualized as a Stackelberg game. The upper-level saves rescue time to determine the transportation scheduling scheme from cities that provide assistance to disaster-stricken cities. The lower-level minimizes cost to determine the optimal allocation scheme of emergency resources from cities that provide assistance to emergency resource distribution centers, which can be characterized as a Nash game among the followers. A robust optimization model with budget constraints is constructed to cope with the demand uncertainty at disaster points, disruption risk at distribution centers and transportation time uncertainty from the emergency resource distribution centers to the disaster-stricken cities. An efficient emergency resource allocation scheme under multiple uncertainties ensures rapid and effective post-disaster recovery, improves disaster supply chain sustainability and reduces environmental pollution. Since the developed model is in a bi-level nonlinear format, the Karush-Kuhn-Tucker conditions and the big-M method are applied to obtain a single-level mixed integer linear programming model. To validate the efficacy of the model, a series of numerical experiments with sensitivity analysis are conducted. Compared with the leader's perspective model and the follower's perspective model, although the computational time of the bi-level model is relatively longer, it can obtain a satisfactory point for both parties. Moreover, as the disturbance of the three uncertainties continues increasing, it can be observed that the two objectives do not necessarily have a simultaneous increasing trend in any case. In addition, under the influence of uncertainty, the costs of cities that provide similar quality assistance can remain stable.