Atmospheric water vapor increases as air temperature rises, which causes further warming. Thus, understanding the underlying causes of atmospheric water vapor change is vital in climate change research. Here, we conducted detection and attribution analyses of atmospheric precipitable water (PW) changes from 1973-2012 over China using an optimal fingerprinting method by comparing the homogenized radiosonde humidity data with CMIP5 model simulations. Results show that the increase in water vapor can be largely attributed to human activities. The effect of anthropogenic forcing (ANT) can be robustly detected and separated from the response to the natural external forcing (NAT) in the two-signal analysis. The moistening attributable to the ANT forcing explains most of the observed PW increase, while the NAT forcing leads to small moistening. GHGs are the primary moistening contributor responsible for the anthropogenic climate change, and the effect of GHGs can be also clearly detected and successfully attributed to the observed PW increases in a three-signal analysis. The scaling factor is used to adjust the CMIP5 model-projected PW changes over China and the observation-constrained future projections suggest that atmospheric water vapor may increase faster (slower) than that revealed by the raw simulations over whole (eastern) China.