Retrieving atmospheric water vapor from remote sensing data is very important in such aspects as weather

forecast, atmosphere correction and the study of climatic change and water circulation. At present, there are the

two-band ratio method and the three-band ratio method for retrieving atmospheric water vapor from near infrared

data. Based on the near infrared radiative transfer theory, this paper has proposed an improved three-band ratio

method under the simulation of the radiative transfer model, MODTRAN. Applying this method to three water absorbing

bands of MODIS, the authors obtained three kinds of water vapor values. Then the method for retrieving water vapor

for MODIS data was worked out on the basis of different sensitivities of the three bands to the absorbing of water

vapor. The authors retrieved the water vapor of eastern China and the results show that the method is feasible and

practical.