Hydrology Lab

Severe rainfall disasters have struck Japan with increasing frequency in recent years. Kyushu in particular has experienced a series of catastrophic events, including the July 2017 Northern Kyushu Heavy Rainfall and the July 2020 heavy rain event in the Kuma River basin, while localized downpours causing fluvial and pluvial flooding now occur almost every year. Combining diverse observations—from satellites, weather radar, and ground stations—with high-resolution regional atmospheric simulations, our laboratory investigates the physical mechanisms that drive extreme precipitation. To support climate change adaptation, we also project local-scale heavy rainfall through dynamical downscaling of future climate simulations and estimate basin-scale Probable Maximum Precipitation (PMP) using regional atmospheric models. Through this work, we aim to strengthen disaster preparedness and contribute to resilient, sustainable communities.

1. Observational Analysis of Heavy Rainfall Mechanisms

High-Resolution Weather Radar (XRAIN)

Japan's Ministry of Land, Infrastructure, Transport and Tourism (MLIT) operates XRAIN, a nationwide network of X-band polarimetric (MP) radars integrated with C-band radars. XRAIN captures surface precipitation across Japan at 1-minute intervals and roughly 250 m spatial resolution, while also providing three-dimensional precipitation structures every 5 minutes. Using these high-resolution radar observations, we analyze the fine-scale structure and evolution of heavy rainfall systems, with a particular focus on northern Kyushu.

Himawari Geostationary Meteorological Satellite

While visible and infrared imagery from the Himawari geostationary satellite is familiar from daily weather broadcasts, its Advanced Himawari Imager (AHI) observes across 16 spectral bands. Leveraging these multi-band satellite observations, we investigate cloud development and atmospheric conditions associated with extreme rainfall events.

Three-dimensional radar analysis of heavy rainfall
(based on XRAIN data and GSI aerial imagery)

2. Numerical Investigation Using Regional Atmospheric Models

Hindcast Simulations of Extreme Rainfall Events

We first reconstruct historical heavy rainfall events through high-resolution, three-dimensional numerical simulations with regional atmospheric models. For extreme events in Kyushu, we configure nested domains down to grid spacings of several hundred meters to 1 km and carefully tune model physics and configurations to faithfully reproduce observed rainfall distributions.

Physical Mechanisms of Heavy Rainfall

Because regional atmospheric models explicitly simulate three-dimensional physical processes, their outputs provide a complete picture of atmospheric fields such as water vapor transport, wind, and temperature. Running simulations at sub-kilometer to 1 km resolution enables us to resolve complex local terrain and diagnose the physical mechanisms that initiate, organize, and sustain extreme rainfall—ultimately helping improve forecasting skill and flood risk reduction.

Numerical analysis of the July 2017 heavy rainfall (Ishida et al., 2017)

3. High-Resolution Projections of Future Extreme Rainfall

Dynamical Downscaling of Global Climate Projections

Global Climate Models (GCMs) used for future climate projections typically operate at horizontal grid spacings of 100 km or coarser due to computational constraints—far too coarse for basin-scale impact studies or for resolving the mesoscale atmospheric processes behind heavy rainfall in Japan. To bridge this scale gap, we dynamically downscale GCM projections using regional atmospheric models.

Nationwide 2 km Grid Simulations Across Japan

We are conducting dynamical downscaling across all Class A river basins in Japan. After dividing the country into multiple computational domains and calibrating the regional atmospheric model for the watersheds within each domain, we downscale future climate projections to a 2 km horizontal grid—capturing the influence of Japan's rugged terrain on future extreme precipitation.

Computational domains covering all Class A river basins in Japan

4. Estimating Physical Upper Bounds of Precipitation

Physically Based Probable Maximum Precipitation (PMP)

Quantifying the physical upper bound of precipitation in a given watershed—known as Probable Maximum Precipitation (PMP)—is vital for flood control planning and the design of major hydraulic infrastructure. Conventional statistical PMP methods struggle to account rigorously for complex terrain or a warming climate. To overcome these limitations, we develop physically based PMP estimation methods using regional atmospheric models.

Application to Watersheds in California, USA

Originally developed and applied across watersheds in California on the U.S. West Coast, this physically based PMP framework earned wide recognition in the United States and received the 2016 J. James R. Croes Medal from the American Society of Civil Engineers (ASCE).

Physically based PMP estimation
(Ishida et al., 2017)