Taiki ONO

Evaluation of Wide-Area Ground Precipitation Estimation Based on Backward Trajectory Analysis of Snowfall

Toshiro KUMAKURA

 The Hokuriku region, including Niigata Prefecture, is known as one of Japan's heavy snowfall areas. Weather radar is commonly used for snowfall forecasting, and in recent years, radar-based precipitation distribution has improved in resolution, allowing for the observation of localized precipitation. However, snowfall during winter is highly influenced by wind while falling from the upper atmosphere to the ground, meaning that the precipitation observed directly below the radar coordinates does not necessarily match the actual ground precipitation. To accurately understand the ground precipitation distribution, it is essential to examine how upper atmospheric meteorological information corresponds to ground-level observations.In this study, we conducted backward trajectory analysis of snowflakes using the 3D wind field from NHM and the Gidai radar, targeting a 10 km �~ 10 km area around the Nagaoka AMeDAS station with a 500 m mesh. We then created high-resolution precipitation distributions. Furthermore, during the period when the correspondence between ground precipitation and 3�� elevation angle radar precipitation was good (December 18, 2022, 18:00 to December 19, 02:00), we treated the 3�� elevation angle radar precipitation as estimated ground precipitation and calculated logarithmic errors at 2-minute intervals for other elevation angles. Based on this, we calculated the standard deviation of logarithmic errors for each distance range and derived the 50% confidence interval.
 The results showed that in periods when the correspondence between analyzed and observed precipitation was strong, the logarithmic error data for each distance range exhibited normality in many cases, confirming that standard deviation could be used for reliability evaluation. After December 18, 23:00, the wind gradually weakened, which may have led to greater displacement in the precipitation distributions at each elevation angle. The difference between ground precipitation and 3�� elevation angle precipitation increased to 1.172 mm/h at 00:00, 0.456 mm/h at 01:00, and 0.346 mm/h at 02:00 on December 19. By averaging the width of the 50% confidence interval CI50_high - CI50_low during the period after 23:00 on December 18 and using it as a threshold, we classified 167 out of 240 cases 69.58% as reliable. This suggests that, overall, the evaluation period set for reliability assessment can be considered valid.