The method of Computational Fluid Dynamics (CFD) is implemented to obtain radiation errors of a temperature sensor equipped with a 076B aspirated radiation shield under various environmental conditions. Then, a radiation error correction equation is obtained by fitting the CFD results using a neural net algorithm. The results show that the error of the radiation error correction equation is within 0.005 ℃. A high-accuracy air temperature data can be obtained by using this correction equation proposed in this paper, which may serve not only for general climate change research but also for thehigh-accuracy weather forecast.