Operational Rainfall Estimation by X-band MP Radar Network in MLIT, Japan

Flood disasters caused by localized heavy rainfall frequently occurred in Japan during July and August, 2008. National Research Institute for Earth Science and Disaster Prevention's (NIED's) results on monitoring of the heavy rainfall and severe wind by using X-band multi-parameter (MP) radar boosted Ministry of Land, Infrastructure, Transport and Tourism (MLIT), Japan to start deploying the X-band MP radars around major urban areas and potentially hazardous areas in Japan. They deployed 11 radars in Tokyo, Nagoya, Kinki, Hokuriku areas in FY 2009, and also deployed 15 radars in other areas in FY 2010. NIED developed an operational data processing system, which estimates the rainfall intensity from the radar data, and which creates regional composite data every one minute, under a commission from National Institute for Land and Infrastructure Management (NILIM), MLIT.

Because MLIT specified the one-minute update of rainfall information, and also needed three dimensional distribution of rain, a little complicated scan strategy is performed in this radar network. Each radar takes three tilt data every one minute. The first tilt is used to create the regional composite data of the rainfall intensity. Two elevation angles are set for the first tilt, and the angle is altered minute by minute, because the regional composite data are created from two tilts to cover the lower tilt shadow area by the higher tilt. (The composite data is the result of two minutes observation, and is updated every minute.) The second and third tilts are a part of a 5 minute volume scan which consists of 12 tilts.

The observed radar data are transferred to two data processing centers in Kanto and Kinki Regional Development Bureaus, MLIT, then the same processing is done for redundancy in these two centers. After quality controls of the received radar data are performed (S/N Ratio thresholding, ground clutter removal, Φ_DP dealiasing, thresholding of Φ_DP deviation from its moving average, ρ_HV thresholding, beam blockage correction, and point clutter removal), K_DP is estimated by a differentiation of Φ_DP, which is filtered by two different FIR filter for removing the local phase shift by strong backscattering and smoothing. As radio wave attenuation by rainfall is not negligible in X-band, the attenuation is estimated from the K_DP to correct Z_H and Z_DR. The estimated attenuation is also used for determining radio wave extinction areas, where the radar cannot detect the received power from the precipitation by the rainfall attenuation. Finally, the precipitation intensity is estimated with Z-R and K_DP-R relationships. The estimated rainfall data are immediately sent to compositing server. The data of radars which belong to the sam1e area are composited into a horizontal mesh with a resolution of about 250 m. For this compositing, the modified Cressman interpolation method, which also considers with a weighting function that becomes greater with the decrease of the height above the ground, is used. These procedures are performed with parallel computing to finish the calculations in several seconds.

The rainfall information calculated by this system is provided experimentally by MLIT since July 2010. Now the validation of the estimated rainfall intensity is being performed with rain gauge and disdrometer data. The accurate information updated every minute is sure to contribute to the monitoring of extreme weather in urban areas.