Flood disasters in river basins often occur due to high rainfall, poor river management, and poor land use. To minimize the impact, this study designed a LoRa (Long Range)-based early warning system in watersheds for efficient long-distance data transmission and reception. This system integrates ultrasonic sensors to monitor water level, Arduino Uno microcontrollers and NodeMCU ESP8266 for data processing, LoRa as a communication medium, and a web server for real-time data display, on the web server the water level is divided into 3 statuses, namely Safe (0-20CM), Beware (21-30CM), and Danger (>31CM). An email-based early warning feature is also used to alert the Hazard status when the water level >31 CM. Tests show that this system has a high level of accuracy in water level readings of up to 50 cm and data transmission up to a distance of 1,000 meters in optimal conditions. With the efficiency of communication and real-time analysis, this system can improve community preparedness for flood disasters.