214 / 2024-02-28 09:01:31

Real-time prediction of river water level based on radar observation

flood,prediction,water level,radar

The risk of river floods becomes more and more serious in the world due to global warming. Therefore, water level prediction in river systems is crucial for flood management and evacuation planning. In conventional methods for water level prediction, the amount of precipitation is usually obtained from pointwise data of rain gauges. On the other hand, the widely spreading weather radars can observe the precipitation all over the river basins. It is very important to develop new methods for the effective use of the observation data. The present work aims to propose new prediction method for river water level and to investigate the validity of the method.

In this study, the Kosui River in Kochi City of Kochi Prefecture in Japan is selected as the test field. As Kochi City frequently experiences heavy rainstorms with precipitation intensity more than 50 mm per hour, floods sometimes occur in the river basin and its surrounding area. We have established a high-resolution X-band polarimetric weather radar network, which observes the precipitation every one minute, covering almost the whole area of Kochi Prefecture including the test field. A prediction equation is derived by assuming a linear relationship between the initial water level and the change in the water level due to rain and drain. The change in the water level is estimated from the information of the radar estimated precipitation and the drainage capacity of the river. In practice, the test field is divided into small grids. The radar estimated precipitation is calculated from R-Kdp relationship at each small grid and considering the infiltration rate into soils and the traveling time to the water level gauge from each grid. For validation, the method is applied to predict the water level at specified location for three heavy rainstorm events. The results show that our method can almost predict the water level except for the slight delay of peak values.