add a new control variable

I have managed to add "W (vertical velocity)" as a new control variable by modifying the codes.  

I used GSI in pure EnVar (beta_s0=0, which means no weight for the static part) and without assimilation of W observation, then I found out, comparing the experiment with and without W as a control variable, the only difference in the "stdout" is that there are increments in W.

But I expect that adding W as a new control variable will have impacts on other variables through the covariance matrix A (related to the ensemble).

Could anyone have ideas why other variables are not affected?

Thanks,

Jia