Study On Coupled Global Atmosphere-land Assimilation Of Near-surface Observations

Near-surface meteorological conditions are closely related to human activities and people's life,which is the main focus of weather forecasting.The global meteorological observation network and the data exchange system were established under the coordination of the World Meteorological Organization(WMO).Observations from more than 10,000 Surface synoptic observations(SYNOP)stations and 5,000 METeorological Aerodrome Reports(METAR)stations are shared globally through Global Telecommunication System(GTS).The assimilation of these observations in global Numerical Weather Prediction(NWP)systems has important scientific and application value.However,the global atmospheric data assimilation system generally only assimilates surface pressure observations,while the observations such as temperature and humidity are used indirectly through the land surface assimilation system(e.g.the European Centre for Medium-Range Weather Forecasts,ECMWF).Only a few NWP centers(e.g.the UK Met Office,UKMO)use near-surface temperature,humidity and wind field observations in their global atmospheric assimilation system.The successful assimilation of near surface observations in UKMO and ECMWF's operational NWP system indicates that,the key point of assimilating near surface observations is to jointly update the global atmosphere and land initial conditions.This paper proposed a new scheme to assimilate near surface observations into Global Forecast System/Grid Statistical Interpolation(GFS/GSI)global cycling data assimilation system.The scheme includes mainly three steps: 1)combine 2m temperature(T2m),2m humidity(Q2m)and 10 m wind(U10 and V10)from the model outputs with observations using Local Ensemble Transform Kalman Filter(LETKF)to produce screen-level analysis;2)drive the global Land Surface Model(LSM)using the screen-level analysis,and assimilate soil moisture in-situ observations using Ensemble Optimal Interpolation(En OI)to obtain the global land surface analysis;3)apply the screen-level observation increments of T2 m,Q2m,U10 and V10 to the lowest level of atmosphere model,and spread the information to upper levels using a hybrid four-dimensional ensemble-variational(Hybrid-4DEn Var)to produce atmosphere analysis.A global atmosphere-land weakly coupled system for assimilating near-surface observations was established based on the proposed scheme.The system is composed of three parts:(1)The screen-level analysis system was established using LETKF.The analysis increments are more flow-dependent compared with OI technique.By the fusion of near-surface temperature,humidity and wind observations,the biases of screen-level temperature,humidity and wind analysis were reduced by more than 50%,while the root mean square root error(RMSE)was reduced by more than 25%.(2)Information of T2 m,Q2m,U10 and V10 observations were firstly fused to screen-level analysis and then transferred to land surface by driving the land surface model.A two-dimensional localized En OIscheme for assimilating in-situ soil moisture observations was also developed to produce more accurate land surface conditions.The long-term assimilation experiments conducted from May to September 2016 shows that assimilating the soil moisture in-situ observations can significantly reduce the bias and RMSE of soil moisture simulations.(3)The idea of applying the observation increment to the lowest layer of atmospheric model and spread the information to upper air by using(Hybrid-4DEn Var technique was proposed and realized in GSIsystem.The single observation assimilation experiment shows that the analysis increment of Hybrid-4DEn Var is more reasonable than three-dimentional variational(3DVar)technique in horizontal and vertical structure and is more flow-dependent.More than 10-days cycling assimilation experiments were conducted using the newly developed atmosphere-land weakly coupled system for near-surface observation assimilation.The results show that the assimilation of near-surface observations can significantly improve the 0h forecast quality of T2 m,Q2m,U10 and V10.The global average RMSE of temperature,humidity and wind were decrease by 17.4%,25.5% and 17.6% respectively.For the 12 h forecast,the RMSE reductions of temperature and humidity are about 3.2% and10.2%,respectively.While the RMSE reduction of wind is about 1%.It is also found that,the near-surface observation assimilation also improves the vertical profile of short-range temperature forecast below 500 h Pa.Based on the GFS/GSI global cycling data assimilation system,this paper constructed a global atmosphere-land weakly coupled system with the ability to assimilate near-surface air temperature,humidity and wind observations.