On the Negative Water Vapor in the NCEP GFS: Sources and Solution

Negative values of water vapor exist in the current NCEP operational global weather forecast model, the Global Forecast System (GFS). Two processes are identified to be responsible for this defect. One is spectral transform from spectral space to grid points. The other is vertical advection of tracers. The contribution from the former is negligibly small and is inevitable in all spectral models. This study is aimed to eliminate the negative source from the later. The current GFS uses central differencing in space and leap-frog in time for vertical advection of traces. The scheme is not positively definite and may produce negative tracers. A few positive-definite schemes are considered and the Van Leer Limiter scheme is chosen. A new time filter is designed to assure computational instability. Both weather forecast experiments and climate simulations show that this new scheme is effective in removing negative water vapors from the GFS. It exerts little impact on the weather forecast skill scores.