Blending Technology of Radar Extrapolation and Mesoscale Numerical Prediction Based on Python

In order to overcome the shortcomings of the current meso-scale numerical models in the short-term prediction of convective-scale quantitative precipitation, and to make up for the shortcomings of the proximity forecasting technique based on "extrapolation" in the ability of quantitative precipitation forecast more than 2 hours. The (QPF) fusion scheme of quantitative precipitation forecast based on "extrapolation" proximity prediction technique and meso-scale numerical model is introduced by China Meteorological Administration, and its experimental application is carried out. First of all, the scheme is based on the (QPE) results of quantitative precipitation estimation based on radar detection and automatic weather station observation. The phase correction of the quantitative precipitation forecast outputted by the meso-scale numerical model is carried out in the spectral space by using the fast Fourier transform (FFT) and the multi-scale optical rheological method to ensure that the global and local displacement deviation of the rain belt is corrected. Secondly, the distribution characteristics of quantitative precipitation estimation are statistically analyzed, and the intensity of numerical forecast precipitation field in the corresponding period is adjusted by approximating the model quantitative precipitation forecast to the quantitative precipitation estimation. Finally, the hyperbolic tangent weight function is used to integrate the corrected numerical model quantitative precipitation forecast with the quantitative precipitation forecast based on proximity forecast technology. The two ends of the tangent curve are based on the weather type of precipitation and the weather change experience of the forecaster. Combined with the temporal and spatial scales of different precipitation systems, different weights are selected in different cases. The forecast experiments and tests of six typical heavy precipitation cases from 2016 to 2018 show that the quantitative precipitation forecast after fusion mainly depends on the "extrapolated" near forecast results in the first 2 hours, and then with the change of fusion weight. The contribution of numerical prediction to the fusion results gradually increased until 5 to 6 hours after fusion. The quantitative precipitation forecast results of 0-6 hours after fusion are improved obviously, which is better than that of single proximity forecast technique or meso-scale numerical forecast model.