This study aims to simultaneously achieve watershed-scale flood management and integrated flood management for both external and internal water systems by:

1. Utilizing advanced technologies (e.g., AI, big data, remote sensing) to produce accurate and precise hydrological and hydraulic information across entire watersheds, including ungauged areas;

2. Applying this information to the development of high-resolution distributed hydrological models, a supercomputer-based spatiotemporal flood inundation information system, and an operational system for runoff reduction facilities that link internal and external water systems in adjacent watersheds; and

3. Developing optimal flood volume allocation techniques for dams and rivers, as well as optimal operational strategies for runoff reduction facilities, to minimize flood risks caused by both external flooding and internal water inundation.

The primary objective of this study is to develop a deep neural network-based high-resolution precipitation forecasting model with accuracy sufficient for practical applications in water resource management, such as mitigating flood and drought risks. To achieve this, the study focuses on: 

(1) developing an algorithm to enhance the accuracy of spatiotemporal gridded precipitation observation data based on radar reflectivity; and 

(2) devising an ensemble technique that integrates multiple nowcasting models specialized in estimating precipitation within specific ranges.