Abstract: The technique of retrieving precipitable water vapor (PWV) based on near-infrared (NIR) data of Medium Resolution Spectral Imager (MERSI) on board FY-3A satellite is introduced. Five NIR channels are designed on the MERSI instrument for PWV observation, three of which are water vapor absorption channels centered near 905 nm, 940 nm and 980 nm respectively and others are atmospheric window channels at 865 nm and 1030 nm. The method adopted here for PWV retrieval is based on the ratio of reflected solar radiance (or apparent reflectance) detected by satellite between water vapor absorption channels and atmospheric window channels. By employing channel ratios, the aerosol extinction distribution and the variation effect of surface reflectance are partially removed, and the atmospheric transmittance of water vapor channels is approximately obtained. The PWV is derived from the atmospheric transmittance based on a Look-up Table which is pre-calculated using a radiation transfer model. The sensitivities of atmospheric transmission in each NIR water vapor channels of MERSI to the total precipitable water vapor are also simulated. It is found that 905 nm channel is more sensitive under humid conditions while the strong absorption channel at 940 nm is sensitive under dry conditions. And the two weak absorption channels have similar sensitivity to total water vapor amount. In this case, under a given atmosphere condition, the derived PWV values from three water vapor channels may be a little different. The weighted average of three derived PWV values is regarded as the final PWV product and the weighing coefficients are determined by their sensitivity.The procedure of the operational PWV product generation is designed and conducted for experimental retrieval. Based on the global data of MERSI, FY-3A Products Generation System (PGS) can successfully generate the daily global and regional PWV L2 products and multi-day integrated L3 products, which can clearly display the spatial distribution of water vapor amounts over global land area. The result indicates that FY-3A/MERSI has an excellent ability in detecting NIR water vapor, and can demonstrate fine characteristic of PWV spatial distributions. As 940 nm channel shows good application under dry atmosphere conditions and 905 nm or 980 nm channel work well under humid situation, acceptable retrieval accuracy can always be achieved by combining these channels. In order to assess the accuracy, the retrieved PWV from MERSI NIR are compared with the ground-based sounding data. Over cloud free area, there is a good agreement between them in variation trend and spatial distribution. The MERSI PWV results are steady but 20%—30% lower than sounding, so the retrieval algorithm and the Look-up Table need to be updated to reduce this bias in the near future.