Weather Forecasting in the Tropics with Climate Models Is Feasible

Boyle, J., Lawrence Livermore National Laboratory

General Circulation and Single Column Models/Parameterizations

Cloud Modeling

Climate Model Forecast Experiments for TOGA-COARE. J. Boyle,S. Klein,G. Zhang,S. Xie,X. Wei. Accepted by Monthly Weather Review


Figure 1. Profiles of the apparent heat source (Q1) at the TOGA-COARE central site for the observations and day-two forecasts of the CAM, CAM with Zhang modification (ZMO), and AM2 averaged over the entire TOGA-COARE period are shown. Units are degrees Kelvin day. The CAM with the Zhang modified deep convection produced the best fit to the observations. Proper simulation of both the magnitude and level of maximum heating were shown to be important for capturing the MJO.


Figure 2. Hovmoller diagrams of the band pass filtered (30-70 days) 200 hPa velocity potential averaged from 5N to 5S for the TOGA-COARE period ( 1 Nov 1992 - 28 Feb 1993). This diagram isolates the MJO signal. The top figure shows the observations (ERA-40). The two MJOs are seen propagating west to east. The model figures are concatenations of day-six forecasts. As shown, the CAM is unable to maintain the MJO signal for the six days. The Zhang modified CAM and AM2 both provide good forecasts of this field with the Zhang CAM modification being a bit too strong.


Figure 1. Profiles of the apparent heat source (Q1) at the TOGA-COARE central site for the observations and day-two forecasts of the CAM, CAM with Zhang modification (ZMO), and AM2 averaged over the entire TOGA-COARE period are shown. Units are degrees Kelvin day. The CAM with the Zhang modified deep convection produced the best fit to the observations. Proper simulation of both the magnitude and level of maximum heating were shown to be important for capturing the MJO.

Figure 2. Hovmoller diagrams of the band pass filtered (30-70 days) 200 hPa velocity potential averaged from 5N to 5S for the TOGA-COARE period ( 1 Nov 1992 - 28 Feb 1993). This diagram isolates the MJO signal. The top figure shows the observations (ERA-40). The two MJOs are seen propagating west to east. The model figures are concatenations of day-six forecasts. As shown, the CAM is unable to maintain the MJO signal for the six days. The Zhang modified CAM and AM2 both provide good forecasts of this field with the Zhang CAM modification being a bit too strong.

Climate models are traditionally evaluated statistically. For example, the monthly means of surface temperature are compared between the models and observations. In such an evaluation, errors from different sources are mixed making a specific diagnosis of model shortcomings difficult. On the other hand, the forecasts of numerical weather prediction models are evaluated for specific sequences of weather events. If sufficient data are available, the workings of the forecast model parameterizations can be compared precisely to the observed processes.

In this work, the National Center for Atmospheric Research (NCAR)/U.S. Department of Energy (DOE) Community Atmospheric Model (CAM) and the Geophysical Fluid Dynamics Laboratory (GFDL) Atmospheric Model 2 (AM2) are cast into the role of forecast model. The models are initialized from observations for a specific time, and a forecast of ten days is generated. The period of interest here is the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA-COARE) from November 1992 through February 1993. The core to this experiment was a comprehensive suite of observing systems set up in the Tropical Western Pacific centered at 2S and 155E. In this period, two well-defined examples of the Madden Julian Oscillation (MJO) passed through the observation array. The current climate global climate models (GCMs) do not depict this oscillation well, and the proper simulation of the MJO is vital to adequately capture tropical variability. The control CAM did not simulate the MJOs. However, the AM2 and CAM with a modified convective parameterization performed somewhat better. The CAM modification was developed by ARM-sponsored scientist G. Zhang. Data from the ARM Climate Research Facility's Southern Great Plains site were important for the development of Zhang's new scheme. The intensive observations and the model forecasts were used together to gain insight into the specific deficiencies of the deep convection parameterizations. This same methodology will be applied to data from the Tropical Warm Pool-International Cloud Experiment (TWP-ICE), conducted over a four-week period in January and February 2006. It is expected that the detailed observations of TWP-ICE will yield valuable information on the cloud processes simulated by the models.

Forecasts using climate models in conjunction with observational data are able to accurately describe the atmospheric processes and can pinpoint problems with model parameterizations. Such problems can become blurred or mixed with other errors in the traditional statistical climate model evaluation.