The Accuracy Analysis Of Data Measured By Ground-based Microwave Radiometer And Assimilation Experiment

The temperature profile at 00 and 12 UTC measured by radiosonde at Nanjiao in Benjing is used to be a standard to assess the data accuracy measured by ground-based microwave radiometer and data simulated by mesoscale model-WRF (The Weather Research & Forecasting Model). The result by comparing three kinds of profiles at 00 and 12 UTC does tell us something important. First of all, mean bias between those those three profiles have same figure, such as they all have maximum mean bias in summer and minimum mean bias in winter. Second, the data retrieved by ground-based microwave radiometer seems much closer to radiosonde than simulated profile just in lower 1 km, but upper 1 km, things totally changed that the data profile simulated by WRF model is closer to sounding data which we treat as real atmosphere profile data. Besides, what should be pointed out is that inversion layer’s existence and its strength have a big influence on temperature profile retrieving of ground-based microwave radiometer, which makes bias profile between ground-based microwave radiometer and rediosonde temperature significantly larger at 00 UTC than 12 UTC. The second step, we use MonoRTM simulation brightness temperature and radiosonde observation data between 2011 and 2013 to train BP neural network, and use this network, observation brightness temperature after deviation correction as network input, to retrieve temperature,relative humidity and water vapor dentisy. The result shows us that deviation correction and BP neural network has a good feedback closing radiosonde obaervation. The third step we put the BP network retrieving temperature and relative humidity into GSI assimilation system, and try to get WRF 24 forecast result for rain,temperature and relative humidity. It tell us WRF forecast precipitation is much weaker than TRMM observation, but temperature and relative humidity bias is not so far.