Comparative experiments on two high spatiotemporal resolution blending algorithms for quantitative precipitation nowcasting

For a long time, the radar echo extrapolation method is the main technique applied for the 0-2 h nowcasting system. However, its actual effective lead time is only ≤ 1 h. The mesoscale numerical model is limited by the spin-up time and the first two hours of precipitation prediction is invalid. In order to solve the deficiencies in the above two types of prediction methods, the most popular technique that blends the extrapolated prediction and the numerical model prediction is applied worldwide to develop uniform 0-6 h lattice high resolution seamless quantitative precipitation prediction system. In this paper, two high spatiotemporal resolution blending algorithms for quantitative precipitation prediction are compared to provide references for their operational applications. In the RAPIDS algorithm, the precipitation intensity and phase of the model prediction are corrected based on quantitative precipitation analysis of the automatic station rainfall merged with radar precipitation estimation. In the INCA algorithm, the extrapolation of the precipitation movement vector is corrected by the wind field of the numerical model. In the two methods, the weight of the extrapolated prediction is gradually reduced and the weight of the model prediction is gradually increased within the lead time of 0-6 h, achieving a smooth transition between the extrapolated prediction and the model prediction. The experimental results show that the two methods are superior to a single extrapolation prediction or model prediction for rain band and precipitation intensity forecast. The development of an optimal seamless blending algorithm for precipitation prediction with high spatial and temporal resolution, which combines the advantages of the two methods, will help to further improve the high resolution quantitative 0-6 h seamless precipitation prediction.