28IOAS NWP-Based Evaluation of Radiosonde Network and AMDAR Data in China

In order to discuss three key questions, including: (1) Where is the key area to deploy the upper-air observation for improving the regional NWP? (2) Whether the AMDAR data can be used as an alternative to the radiosonde data at some time or in some region? (3) Is there any cost-effective way to improve the operational sounding frequency to better capture the variation of high impact weather? A series of experiments are conducted.

The sensitivity areas and levels are analysed by using Forecast Sensitivity to Observation Impact (FSOI) method and Ensemble Transform Sensitivity Method (ETS) method. FSOI experiment with 15km resolution and 30 vertical layers assimilates the simulated Radiosonde observation which extracted from ERA-5 analysis data and superimposed normal random noise (average=0.0, Standard deviation comes from WRFDA system). The WRF model and WRFDA system version 3.8 is utilized for the FSOI experiment. GRAPES model and 0.5°*0.5° GRAPES-GLOABAL ensemble forecast are utilized for the ETS experiment. The verification area domain of both two experiments is eastern China (106.5°E-124.3°E,23.7°N-38.7°N). Both experiments results show that the upper-air observation, especially from 500hPa to 200hPa on the plateau and the middle and lower reaches of the Yangtze River is very important for 12h forecast of eastern China.

Then, the possibility of AMDAR data replacing sounding in some areas or some time is researched by Observation System Experiment (OSE) method with WRF model and WRFDA system version 3.8. The OSE experiments consists of four scheme: using all radiosonde observation, using radiosonde observation except 00UTC and/or 12UTC 11 radiosonde which nearby the airports. a variety of observation are also absorbed in all experiments, including AMDAR, Radar velocity and reflectivity, SYNOP, wind profilers. It can be find that no significant influence for analysis and 12 hour forecast if remove the 11 sounding nearby the airports, but under current situation of network, radiosonde could not be replaced by AMDAR at lower layer on the national scale.

In addition, the frequency experiment of radiosonde release is executed by an observing system simulation experiment (OSSE) approach. The NCEP FNL data is used as nature-run for constructing simulated sounding observations at 06 UTC and the GRAPES-Meso model is utilized for the numerical experiments. The result of OSSE shows that the simulated sounding observations at 06 UTC have significant influence for forecast last up to 24 hours. Combined with long distance sounding experiment, it cloud be known that 6 hours interval sounding is necessary for intensive weather forecast. A cost-effective stratagem for improving the temporal and spatial resolution of sounding is proposed.