Abstract:The periodical calibration of various meteorological sensors of AWS is very important in order to ensure accuracy, reliability and comparability on th e measuring value. The scene calibration of AWS is different with the laboratory examination. The calibration result has been the obvious uncertainty by the imp act of the objective condition. To ensure accurate and reliable value of a quant ity transfer, the improvement method is proposed so as to reduce misjudgment of calibration results. Based on calibration data in Shandong province from 2004 to 2006, the tendency of calibration error is analyzed with two methods of convent ion and improvement. The results show that calibration errors of temperature, hu midity and pressure sensor have been decreased 0.1℃, 1% and 0.2hPa, respect ivly and the ultra error quantity of ground temperature sensor is dropped from 2 3.6% to 5.2% with new method. In a word, the improvement method can rednce the i mpacts of natural enviroment on calibration results, it can avoid to replace sen sors blindly, so it makes up conventional method.