Abstract: The wireless sensor network (WSN) technology has been evolving very quickly in recent years. Sensors are constantly improved in their sensing, processing, storage, and communication capabilities. In a larger scale hilly land agricultural environment monitor WSN, a pure multi-hop approach to route the data all the way along the network, which can extend for hundreds or even thousands of kilometers, can be very costly from an energy dissipation point of view. In order to significantly reduce the energy consumption used in data transmission and extend the network lifetime, we presented a three-tier framework for monitoring agricultural environment using TUFSN（Three-tire Unmanned aerial vehicle Farmland Sensor Network）where data collection and transmission were done using Unmanned Aerial Vehicles (UAVs) in South China. In the system, we defined three types of nodes, which included: sensor nodes, relay nodes, UAVs. In the sensor nodes, a classic WSN one-hop or multi-hop routing approach to transmit their data to the nearest RN was used, which acted as a cluster head for its surrounding sensor nodes. Then, an UAV was moved along the pre-defined route or optimizing route that transported the data collected by the RNs to the data center. The TUFSN was divided into three layers, which included: the lower layer for acquisition data based on relay node and sensor node, the middle layer of relay transmission based on sensor node and UAV, and the upper layer for moving aggregation based on UAV and data center. This architecture led to considerable savings in node energy consumption, due to a significant reduction of the transmission ranges by use of a one-hop or the least hop between sensor node and relay nodes and the transmission to communicate the data by use of a one-hop from the relay nodes to the UAV. Furthermore, the strategy provided reduced interference between the relay nodes that can be caused by hidden terminal and collision problems, which would be expected if a pure multi-hop approach was used at the relay node level. We evaluated the performance of the architecture by some simulation. Our simulation of investigating the impacts on some parameters included: flying speed, flying height, flying time, relay node buffer size, UAV buffer size and relay node communication radius. Simulation and test results showed that, 1) when flying at the speed of 1 m/s, UAV should fly much lower to get enough time for communication with relay node; 2) when the communication radius of relay node was higher than 30 m, UAV could have more than 45s for communication; 3) when the buffer size of each relay node increased, UAV needed to fly more slow or fly more lower; 4) when relay node communication radius increased, UAV could fly more high or fly more quickly; and 5) when the transition ratio was 2000 bps, the range of relay node's cache was 3-13kB. We hoped that our simulation provided some guidelines on large scale hilly country farmland data collecting systems. A set of experiments were carried out in the Teaching and Research Farm of South China Agricultural University in April 2015, with the purpose of demonstrating the effectiveness of the proposed system. The experiments were limited to eight relay nodes which were separated from each other at a distance of 150 m. There were two or three sensor nodes surround every relay node. It should be highlighted that sensor nodes can only communicate with the closest relay node. The eight rotors of the UAV attached mobile node flied at a speed of 1 m/s and 15 meters altitude in the experiment. Statistical results indicated that the average communication time for UAV and each relay nodes were 26 seconds.