Nonlinear normal mode initialization of a limited area model using nonstationary Rossby modes

Application of the nonlinear normal mode method to the intializationof a limited area prediction model requires the construction of the normal
modes of the linearized model equations. Linearized model equations, in differentiated form, can be derived in different ways, depending on the splitting of the terms which contain the Coriolis parameter f. Instead of setting f equal to a constant value in these equations, leading to stationary Rossby modes, we consider a simple extension, by modifying the linearized model equations so that they admit solutlons with nonzero Rossby frequences.
The results are compared with those using a nonlinear normal mode initialization with constant Corfolis parameter. They show that the two methods are virtually identical in their effect upon the initial fleids and upon their development during the first hours of the forecast.