Faced with a series of limitations in developing numerical weather prediction model, a brand new conception of forecaster data assimilation (DA) has been put forward. That is an organic combination between weather forecasters’ irreplaceable ability and numerical weather prediction. The improvement on the model initial field by manual correction of potential vorticity is coincident with this conception. By utilizing the principle of correspondence between potential vorticity (PV) and water vapor image (WV) observed by satellite, forecasters can exam the model and manually correct the distribution of PV, and use PV inversion to obtain a well developed model initial field. This technology has been proved efficacious in operational work in some countries. This study takes use of the principles mentioned above to achieve the whole technology and use some experimental results to show the technical processes.