Simulation of observed low-frequency anomalies using generalized barotropic models

Using the conservation of potential vorticity, generalized spectral barotropic models have been constructed that apply at a single isentropic level. An effective squared wavenumber has been calculated by performing linear regression of each spherical harmonic component of potential vorticity against the corresponding component of streamfunction. Data from the NCEP reanalysis from 1990 to 1999 have been used. This effective squared wavenumber serves as a spectral generalization of the barotropic squared wavenumber ( n(n+1) ).

For the present study, data from the winter months of 1994 and 1995 are low-pass filtered using a Lanczos filter which retains anomalies with periods greater than 10 days. The model equation is linearized about the upper-level basic flow of the two winters. The model includes linear drag and biharmonic diffusion. For each day in the data record, given the low-frequency streamfunction anomaly, the 10-day model forecast is predicted and compared with the observed 10-day anomaly. To obtain a more succinct overview of the model's skill, the observed 10-day lag covariance of low-pass filtered streamfunction anomalies is compared with the predicted 10-day lag covariance.

It is conjectured that the empirically modified models exhibit increased skill in the simulation of low-pass filtered observed streamfunction anomalies, because of the more accurate representation of the upper-level fluid flow. Also, the use of potential vorticity ensures that baroclinic information is included and thus the equivalent barotropic nature of low-frequency anomalies is incorporated in the models.