The assimilation of round-trip horizontal drift radiosonde data in CMA-MESO 3DVar and its impact on model forecast
The round-trip horizontal drift radiosonde can do the three-stage observation of "ascending-drifting-descending stages" by releasing a sounding balloon at only once time. In particular, the descending stage increases the vertical observation of the atmosphere at 06UTC and 18UTC respectively, which has the potential to significantly improve the prediction skills of the regional high-resolution rapid assimilation cycle prediction system at 06UTC and 18UTC. In order to realize the assimilation of the round-trip horizontal drift sounding in the high-resolution regional model and analyze its impacts on the forecast, one vertical thinning method by ‘selecting the closest observation according to the model layer’for assimilation is preliminarily proposed, and the influence of this thinning method on model analysis is deeply analyzed. On this basis, a one-month batch assimilation impact experiments are carried out by using the networked observations in the middle and lower reaches of the Yangtze River and the impacts on model forecasts are detailed investigated. The results of the thinning sensitivity tests show that compared with the assimilation of traditional operational soundings, the analysis and prediction root mean squared errors (RMSE) of the assimilation of non-thinning round-trip horizontal drift radiosonde observations are significantly increased, and the precipitation prediction scores are also significantly declined. On the contrary, the performance of analysis and prediction fields are improved by assimilating the data after thinning，which indicate that the vertical sparsity of round-trip horizontal drift radiosonde observations must be carried out before assimilation. The results of batch experiments show that at the cold start time (00UTC and 12UTC, time with conventional radiosonde), the assimilation of round-trip horizontal drift sounding (ascending stage) observations has little change in analysis error and prediction error than assimilation of traditional operational radiosonde data. At the warm start time (03UTC, 06UTC, 09UTC, 15UTC, 18UTC, 21UTC, time without conventional radiosonde), compared with the control experiment, the accuracy of the analysis field is improved by about 0.4 % by assimilating the descending section of the horizontal drift sounding data . The ETS scores of the 12-hour cumulative precipitation forecast of 0-12 hours change little, and the ETS scores of the 12-hour cumulative precipitation forecast at 0.1 mm, 1.0 mm, 5.0 mm, 10.0 mm and 25.0 mm threshold increase by about 0.5 %. Moreover, the ETS scores at 50.00 mm threshold increased by about 2.3 %. All above results show that the round-trip horizontal drift sounding must be thinned before assimilation, and the assimilation of the round-trip horizontal drift sounding can improve the precipitation forecast skill of the regional high-resolution rapid assimilation cycle forecast system at the warm start time.