<![CDATA[Floods are among the most frequent hydrometeorological disasters in Indonesia and cause significant impacts on public safety, economic activities, and community infrastructure. To support improved preparedness, a water level monitoring system that operates in real time, provides accurate measurements, and is easy to implement in the field is required. This study aims to develop a prototype of an Internet of Things (IoT)-based water level monitoring and classification system using an ESP32 microcontroller, an HC-SR04 ultrasonic sensor, and a Tiny Machine Learning (TinyML) model that performs on-device inference. Water condition determination is based on predefined criteria and water level thresholds established during the analysis stage, which are subsequently used for data labeling and TinyML model development. The system is equipped with LED indicators (green, yellow, red), a 0.96-inch I2C OLED display for local monitoring, and an automatic notification mechanism via the Telegram application as a remote early warning system. The research adopts a Research and Development (R&D) approach using the ADDIE model, which includes needs analysis, system design, hardware and software development, prototype implementation, and system performance evaluation. Experimental results show that the system is able to measure water levels stably with a maximum error of 0.2 cm. A Mean Absolute Error (MAE) of 0.025 cm was obtained from eight experimental trials within a test range of 0–22.5 cm. The system successfully classified water conditions into “Safe,” “Alert,” and “Flood” categories using the TinyML model, achieving 100% agreement with test scenarios across ten test conditions. Furthermore, the Telegram notification mechanism operated as designed, with warning messages sent during all tested warning status transitions (five out of five scenarios) and not sent under stabel safe conditions. These findings indicate that the integration of IoT and TinyML has strong potential to support responsive and efficient flood monitoring and early warning systems, suitabel for small- to medium-scale prototype deployment in residential areas prone to inundation with measurable water level variations.]]>
