Datas of Thermal infrared channels from both FY-3A Moderate Resolution Imaging Spectroradiometer (MERSI) and Visible and Infrared Radiometer (VIRR) can be used for retrieving Land Surface Temperature(LST). Combination of spectral response function from thermal infrared channels, using MODTRAN to simulate the bright temperature under the conditions of clean air, urban aerosol, rural aerosol and dust aerosol, and calculating LST of different aerosol types and optical thickness based on split window algorithm. The results show that the differences between aerosol types and optical thickness will cause brightness temperature changes in the thermal infrared channel, especially in VIRR. Through the relationship between fitting coefficients and aerosol optical thickness, establishing a new Retrieval algorithm. It’s found that adding split window algorithm with aerosol optical thickness can decrease root mean square error of LST, especially in the aerosol concentration is high, the effect is better. The effect of urban aerosol is improved obviously(RMSE decreases from 0.814k to 0.442k , and bias decreases from 0.998k to 0.668k).