In this paper, the problem of minimizing the age of information (AoI) of drone-based systems is studied. In particular, unmanned aerial vehicles (UAVs) that are equipped with sensors assist a base station (BS) to monitor several physical processes of interest by sampling and transmitting the status information of these physical processes. Due to wireless resource limitations, the BS can only receive the sampled status information from a subset of UAVs and estimate the status of the unsampled physical processes by historical status information. Thus, each UAV must dynamically adjust its sampling policy according to the variations in its monitored physical process, while the BS must determine the subset of UAVs for status information transmission so as to maintain the freshness of the sampled status information. The AoI is used to measure the freshness of the status information. Hence, we formulate this problem as an optimization problem aiming to minimize the overall AoI, including the effects of physical dynamics of the physical processes being monitored. A QMIX-based algorithm is designed to solve this problem. The proposed approach enables the UAVs to optimize the sampling policies with local information and the BS to select the optimal subset of UAVs for status information transmission, thus minimizing the system AoI. Simulation results based on real data of PM2.5 density in Beijing show that the proposed method can reduce the system AoI by up to 62.6% and 75.3% compared to the deep Q-network method and the uniform sampling policy, respectively.