Errors Analysis on Estimating Raindrop Size Distribution in a Mei-Yu Front Heavy Rainfall Process

The gamma function has been accepted widely to model raindrop size distribution (DSD) in recent years. And errors on DSD spectrum in a persistent Meiyu front heavy rainfall process from 30th on June to 3rd on July in 2016 in Hubei province over central China are analyzed with one Particle Size and Velocity (PARSIVEL) in observation station. After estimating simulated and observed rain DSDs, it is shown that: (1)The root mean square errors (RMSEs) between simulated and observed moments are between 0.1 and 0.4. It shows a good agreement for estimators with observation. (2) The RMSEs of number concentration (N) of raindrops increase from the low-moment estimators to the high-moment ones. And the characteristic is not change with rainfall intensity, which may be caused by more weight from little raindrops. So the RMSEs of natural logarithm of N (lnN) from the low-moment estimators to the high-moment ones are analysed. M456 (using the fourth-fifth-sixth moments) produces the largest errors, followed by M012, then by M346 and M246. While M234 produces the least errors. (3) There exist differences between observed and simulated DSDs. The low-moment estimators M012 produces less errors when diameter of raindrop (D) is less than 2 mm, while errors increase with D when it is greater than 2 mm and the contribution of greater raindrops to precipitation is underestimated. The mid- and higher-moment estimators produce less errors than M012 when D < 0.375 mm. And the differences among mid- and higher moment estimators to each other are not too big when D <= 4.25 mm. When D is greater than 4.25 mm, RMSEs between M456 and observed DSDs increase with it and the differences among other three estimators (M234, M246 and M346) are not substantial. It shows that RMSEs of N and lnN could not help us to select the best scheme to represent the real DSDs spectrum. And different estimators have their own advantages. Besides, the slope λ, shape μ and intercept parameters (N0) from different estimators are compared and big difference are found especially in N0.

Key words: Raindrop size distribution; Errors analysis; Meiyu; Heavy rainfall process