Research on a Rainfall Prediction Model in Guizhou Based on Raindrop Spectra

Abstract

Our study and analysis of the distribution differences in raindrop spectra in a Guizhou precipitation prediction model were of great significance for understanding precipitation microphysical processes and improving radar quantitative precipitation prediction. This article selected the Dafang, Majiang, and Luodian stations at different altitudes in Guizhou and analyzed the distribution characteristics of precipitation particles at different altitudes. This article used precipitation data from the new-generation Doppler weather radar, OTT-Parsivel laser raindrop spectrometer, and automatic meteorological observation stations in Guiyang via M-P and GAMMA and established methods to fit the particle size of raindrop spectrum precipitation. Based on the LSTM neural network method, we constructed a precipitation prediction model for Guizhou and conducted performance testing. The results show that (1) the precipitation particles at the three stations are all concentrated in small particle size areas, with a peak value of 0.312 mm and a final falling velocity of 1–5 m/s, and the particle size increases with a decreasing altitude. The contribution rate to the density of particles with a precipitation particle size of less than 1 mm exceeds 80% and decreases with a decreasing altitude. The average volume diameter of precipitation particles has the highest correlation with the precipitation intensity. (2) In the fitting of the raindrop spectrum distribution, the GAMMA distribution fitted by the three stations has a better effect and the fitting effect gradually improves with an increasing altitude. (3) In precipitation prediction for convective clouds and stratiform clouds, the 60 min prediction results are the most consistent with the actual precipitation, with correlation coefficients of 0.9287 and 0.9257, respectively, indicating that the prediction has high reliability.

Keywords

• raindrop size distribution
• microphysical characteristics
• LSMT neural network
• precipitation forecasting