In recent years, severe rainfall events, such as linear precipitation zones and typhoons, have inflicted considerable damage on human lives as well as agricultural and livestock products. Consequently, research is transitioning from traditional kilometer-scale, hourly predictions using rain cloud radars to highly detailed, real-time microclimate forecasts at the meter scale. However, when addressing actual heavy rainfall disasters, especially in mountainous areas, it is essential to consider not only the atmospheric conditions predicted by micrometeorological models but also the rapidly changing geographical conditions, localized groundwater eruptions, reservoirs, rivers, and buildings, which are specific to the region's geography. Therefore, this paper aims to propose a highly detailed and reliable Early Heavy Rain Warning System utilizing cloud-based micrometeorological forecasting and numerous IoT sensors. Specifically, the proposed Enhanced MQTT protocol is introduced to facilitate simultaneous wireless connections of a large number of IoT sensors and agricultural drones for observing geographic factors. By integrating micrometeorological predictions from cloud services with these geographical observations, anomaly detection using the Extended Kalman Filter is proposed for early warning recognition. Finally, the paper presents the simulations for evaluating the proposed methods and discusses future research possibilities.