A Study Of Assimilation Of Surface AWS Data Using WRF-EnSRF

The elevation difference between model surface and observation site is always a problem difficult to handle in surface data assimilation. A reasonable assimilation scheme is able to assimilate surface automatic weather station (AWS) data into Meso-scale model efficiently. In this paper, surface AWS data are assimilated into WRF model through an ensemble Kalman filter firstly using Guo et al.(2002)'s scheme. Then an adjusted scheme is given, assimilating 10-mwind observations,2-mpotential temperature,2-m dewpoint temperature, and surface pressure. Validity is proved by the mean square root error analysis, simulated result and assimilation increment analysis. And sensitive experiments are given for checkout the response of assimilation of each AWS meteorological parameter.For put surface Automatic Weather Station (AWS) data into numerical models sufficiently, a further improvement is proposed to solve the impact of assimilation results caused by elevation difference between observation site and model surface. Potential temperature and dewpoint temperature Terrain Error of Representativeness (TER) are added into temperature and dewpoint temperature error of surface observation data assimilation in WRF-EnSRF system respectively, and a suitable Temperature Lapse Rate (TLR) is determined by test results. The main conclusion includes:(1)The assimilation of surface AWS data through Ensemble Square Root filter using Guo et aL(2002)'s scheme can improve the simulation results. (2)The assimilation of any element of surface observation data(temperature, humidity, wind, surface pressure) separately can affect the forecast of 18 hour accumulated rainfall, but different elements have different impacts, in which the most influenced one is dewpoint temperature. (3)The use of 2-m potential temperature and 2-m dewpoint temperature instead of 2-m temperature and 2-m specific humidity get a better simulation results. (4)The Root-Mean-Square Error (RMSE) only has a little change when TERs are added. (5)Different TLRs make different simulation results of 18 h accumulated rainfall which a suitable TLR is able to improve, compared with no TERs added. (6)When we choose 0.4 K/100m as TLR, we get a better results correspondent with observations.