Precipitable water vapor (PWV) shows great significance in remote sensing quantitative application and ecological research. Aiming at solving the problems of traditional methods, an Improved Continuum Interpolated Band Ratio (ICIBR) algorithm was proposed based on the ratios of apparent reflectance of multi-channels in this paper. The ICIBR algorithm considers the absorption characteristics of water vapor absorption in three MODIS near infrared channels (Bands 17, 18 and 19) and the relationship between the PWV and the ICIBRs of above three channels were simulated by using the MORTRAN model. Then the PWV retrieval model for MODIS data was constructed. Texas, Oklahoma region, a typical arid/semi-arid areas, located in North South America were selected as the study area and four different MODIS 1B data were obtained to perform PWV retrieval experiments using the ICIBR algorithm. Last, the corresponding GPS PWV ground observation data provided by SuomiNet and the MODIS PWV product (MOD05) were obtained to verify the experiment results. Evaluation and comparison results showed that the PWV retrievals showed a higher consistency (r=0.967) with the GPS ground measured PWV data with smaller RMSE~0.276 cm and a total of 71.08% of observation points falling within PWV Expected Errors (EE~±0.05+0.15PWVgps). Moreover, the ICIBR algorithm showed an obviously great improvement in PWV estimation, which can effectively reduce 61% overestimation of PWV retrievals than MOD05 PWV products. This new algorithm is more simple and practical with an overall more reliable retrieval accuracy.