INTERACTIONS BETWEEN MOIST HEATING AND DYNAMICS IN ATMOSPHERIC PREDICTABILITY

The predictability properties of a fired heating version of a GCM in w hich the moist heating is specified beforehand are studied in a series of identical twin experiments. Comparison is made to an identical set of experiments using the control GCM, a five-level R30 version of the COLA GCM. The experiments each contain six ensembles, with a single e nsemble consisting of six 30-day integrations starting from slightly p erturbed Northern Hemisphere wintertime initial conditions. The moist heating from each integration within a single control ensemble was ave raged over the ensemble. This averaged heating (a function of three sp atial dimensions and time) was used as the prespecified heating in eac h member of the corresponding fixed heating ensemble. The errors grow less rapidly in the fixed heating case. The most rapidly growing scale s at small times (global wavenumber 6) have doubling times of 3.2 days compared to 2.4 days for the control experiments. The predictability times for the most energetic scales (global wavenumbers 9-12) are abou t two weeks for the fixed heating experiments, compared to 9 days for the control. The ratio of error energy in the fixed heating to the con trol case falls below 0.5 by day 8, and then gradually increases as th e error growth slows in the control case. The growth of errors is desc ribed in terms of budgets of error kinetic energy (EKE) and error avai lable potential energy (EAPE) developed in terms of global wavenumber n. The diabatic generation of EAPE (G(APE)) is positive in the control case and is dominated by midlatitude heating errors after day 2. The fixed heating G(APE) is negative at all times due to longwave radiativ e cooling. The linearized interactions of the errors with the mean flo w in the thermodynamic equation lead to the creation of EAPE via the t erm C-APE, but the corresponding term in the momentum equation C-EKE l eads to the loss of EKE, in analogy to the life cycle of baroclinic ed dies. (The choice of the mean flow has a noticeable impact only at sho rt times.) The nonlinear terms, consisting of explicit nonlinearities in the error plus residual quasi linear terms, create both EKE and EAP E. At the earliest forecast times the creation of total error energy i s dominated by C-APE at large scales, and by the nonlinear terms at sm all scales. By day 8 in the control case(later in the fixed heating ca se) C-APE dominates both the nonlinearities and the diabatic generatio n at all scales. The ratio C-APE(Control)/C-APE(fixed heating) ranges between 2 and 3. The larger G(APE) and C-APE in the control case allow for a much stronger conversion of EAPE to EKE than in the fixed heati ng case, by a factor of 4. This conversion is the major source of EKE in both sets of experiments, but dominates over the nonlinear creation of EI(E more completely in the control case. The nonlinearities creat e EKE preferentially at the larger scales.