The Three Rivers Source area is located in the southern part of Qinghai Province,China.It is the country's largest nature reserve,and the world's nature reserve with the highest concentration of biodiversity in a high-altitude area.Summer is when the most concentrated precipitation occurs there.Once rainstorm occurs,disasters such as landslides and flash floods very easily occur,which not only pose a serious threat to the safety of local people's lives and property,but also damage the local economic development.Improving the accuracy of numerical weather forecast in the Three Rivers Source area can improve the forecast level of rainstorm and effectively reduce losses due to disaster.Numerical weather forecasting involves initial boundary value problems,and the more accurate the initial field is,the more accurate the forecast results will be.The essence of data assimilation is to study how to process various unconventional observation data with different accuracies and integrate them reasonably with conventional observation data into an organic whole,so as to provide a more accurate initial field for numerical models and achieve the goal of improving the accuracy of model forecasting.Compared to other assimilation data,advantages of satellite data include consistent observation data,wide coverage,high spatiotemporal resolution,and being unaffected by geographical conditions.The Three Rivers Source area has high and complex terrain,vast area,and few meteorological observation stations,thus there are significant problems with the initial value quality in numerical models.However,satellite radiance data has the characteristics of wide coverage and high spatiotemporal resolution,thus it is expected to improve the current situation of insufficient conventional observation data in the Three Rivers Source area.For this reason,it is imperative to carry out research of MHS (microwave humidity sounder) data assimilation in the Three Rivers Source area,which lacks conventional data.The experimental research process was as follows:First,we selected three typical precipitation cases in the Three Rivers Source area:(June 30,2018,July 5,2018,and August 24,2018).Next,we used reanalysis data from NCEP (National Centers for Environmental Prediction) FNL (final operational global analysis),and added ATOVS (Advanced TIROS Operations Vertical Sounder) humidity detector MHS data assimilation.Then we conducted a cyclic assimilation test of three precipitation processes in the Three Rivers Source area based on the WRF (Weather Research and Forecasting) model and its three-dimensional variational assimilation system,to analyze the simulation status of the three events and perform quantitative analysis of the precipitation results.After the assimilation of the three precipitation events with the MHS data,the results show the following:(1) The amount of water vapor predicted increases,which is most obvious and highly consistent with the actual situation in the middle layer,and has good simulation of water vapor and wind,yet poor prediction of temperature at the ground level.(2) The impact of MHS data on precipitation forecasts is mainly reflected in the expansion of precipitation areas and the increase of precipitation forecasts,but it increases the number of empty reporting areas.(3) From the results of the TS (threat score),ETS (equitable threat score),and POD (probability of detection) score,two precipitation tests showed significant improvement.Among them,the TS score of “0630” test improved by 0.05—0.1 for 0.5—10 mm,ETS improved by more than 0.08 for 5 mm,and also improved by a small amount of approximately 10—20 mm,and the POD test improved between 0.5—20 mm.The most obvious improvement was 0.1—0.25 in the range of 0.5—10 mm,while the TS and ETS of the “0824” test improved by more than 0.1 in the range of 10—20 mm,and the POD test improved by 0.1—0.4 in the range of 6—20 mm.The assimilation improved the precipitation forecast,which was particularly evident in the large-threshold precipitation.(4) The assimilation of MHS data did not significantly improve the precipitation forecast of the “0705” experiment,which indicates that assimilation does not always yield positive effects in the results.Therefore,one should not overly trust it when using MHS data.In general,assimilating MHS data can improve the forecast quality.In addition,there remain the following shortcomings in this experiment:The selected assimilation data is not highly varied,and the three selected precipitation cases are all strong precipitation in shortwave troughs,which may not be applicable to precipitation in other weather systems.Therefore,in the future,higher precision satellite data and microwave humidity data from other satellites can be considered to conduct assimilation experiments on various types of precipitation,so as to further verify the assimilation effect of satellite data in the assimilation prediction system of the Three River Source area.