Abstract: Abstract: The initiation and termination time of rainfall occurrences were essential to the rainfall process. In this study, a novel remote monitoring device was designed to accurately detect the termination time of a rainfall in the facility agriculture. This device was mainly composed of the rain sensing module, rain detection circuit, Arduino UNO microcontroller, GPRS DTU module, a serial communication circuit, antenna, and mobile phone application. A total of 16 flumes were evenly spaced parallel on the upper surfaces of two substrates in the rain sensing module. Two HDPE wires were embedded in the flumes to ensure that the wires were separated from each other without connecting. The rain contacted the two wires when the rainfall initialized, where the resistance value between the wires decreased rapidly. When the rainfall terminated, the water on the substrates surface drained from the flumes, where the resistance value between the wires increased gradually. The rain detection circuit converted the changes in resistance values between the wires into the changes in voltage values, then output the rain detection signals. The rainfall process was detected simultaneously through 3 rain sensing modules with the same specification. 3 Kalman filters were selected to simultaneously optimize the process, after reading the rain detection signals by Arduino UNO. A discriminant model was used to determine the initiation and termination signals of rainfall. The detected signals of rainfall were transmitted to the GPRS DUT module through the serial communication circuit, then to the cloud server for storage. A reference time of rainfall was set according to the storage time of signals in the server and the device system. The mobile phone was applied to receive and view the initiation and termination time of rainfall in the period. The test factors were selected, including the rainfall intensity, the distance between the ends of wires, the angle between substrates, and the discrimination period of the discriminant model, in order to determine the optimal parameters for the discriminant model and rain sensor module. The performance of the device was evaluated using the evaluation indexes, such as the misstatement rate of signal discriminant, the success rate of mobile phone reception, and time detection accuracy. The experimental results showed that: 1) The discriminant model with a discrimination period of 10 s had the best discriminant effect; 2) The signal misstatement rate was 1.2%, while the success rate of mobile phone reception was 100%, indicating the device operated reliably; 3) When the distance between ends of wires was 2 mm and the angle between substrates was 120°, the stability and accuracy of the device operation were better, indicating an optimal combination of parameters to detect the light rain. The verification experimental results showed that the detection error range for the initiation and termination time of rainfall were 7-34 and 9-29 s, respectively, and the signal misstatement rate was 5.9%, which met the detection requirements of natural rainfall, showing the device ran well. Therefore, the device can be expected to serve as the remote monitoring of the initiation and termination time of rainfall.