GFS Boundary Condition Frequency

[msullivan-cornell]

I am downscaling GFS .25deg to 9km for a medium to large domain. In the past I've used the .5deg GFS, which only offered time steps of 3-hours, while GFS .25deg offers 1-hour time steps.

I'm wondering, is there a significant advantage to model accuracy by increasing the boundary condition interval to 1-hour vs 3-hours? or should I just stick with 3-hours?

[dcvz]

The answer, as always, is - it depends. Using my definitions of "medium to large" (as continental U.S.) and "significant advantage" then, no, you won't see a significant advantage. But, it depends on the location of the domain and phenomenon you're trying to predict. High-latitude domains with fast flows may benefit from the higher time-frequency.

For a 9km grid, the lateral boundary update frequency will probably be pretty far down the list of the sources of forecast error. But, you could always try the 1-h interval for a case or two and see exactly what difference it makes to the forecast.