Natural disasters can cause large-scale damage to terrestrial base stations, leading to mobile network service disruptions and severely degrading the efficiency of emergency rescue operations. To rapidly restore post-disaster mobile network performance, this paper proposes an optimization method for coordinated air-ground base-station deployment and resource scheduling in emergency mobile communication networks. First, a post-disaster terrestrial base-station damage model is developed, which jointly considers earthquake intensity, base-station distribution, and structural characteristics of base-station buildings, enabling accurate characterization of the survival probability of terrestrial base stations. Meanwhile, a dynamic user aggregation model is designed to capture post-disaster user mobility patterns, yielding accurate post-disaster user distribution features. On this basis, considering practical constraints such as limited emergency equipment availability and finite UAV endurance in disaster areas, the total energy consumption of aerial base stations is minimized by jointly optimizing the transmit power of air-ground base stations, the placement of aerial base stations, and resource allocation strategies. To meet the requirement of rapid post-disaster deployment, the original non-convex mixed-integer programming problem is decomposed into a terrestrial-side subproblem and an aerial-side subproblem, and a two-stage collaborative algorithm is devised by nesting numerical optimization within reinforcement learning. By embedding numerical optimization into the feedback learning process of deep neural networks, the proposed algorithm reduces the complexity of conventional numerical optimization and significantly accelerates reinforcement learning training. Simulation results show that, compared with baseline algorithms, in high-magnitude earthquake scenarios, the proposed method restores communication coverage using only 68% of the total energy consumption while improving the average user rate by approximately 29%.