Tackling Tropical Convection in Climate Models

Zhang, G., University of California, San Diego

General Circulation and Single Column Models/Parameterizations

Cloud Modeling

Zhang, G. J., and H. Wang, 2006. Toward mitigating the double ITCZ problem in NCAR CCSM3, Geophys. Res. Lett., 33, L06709, doi:10.1029/2005GL025229 (23 March 2006).

Figure 1. Climate models commonly suffer from a problem known as the double-ITCZ, which is illustrated here via observed and model-simulated rainfall at the surface. The error is seen in the region circled where, compared to observations (Image A), the original climate model (Image B) produces a second, erroneous equatorial ITCZ band southward from the one found in nature (error indicated by arrow) and too little rain at the Equator. The improved model in the bottom image (C) greatly corrects these common deficiencies. The simulations shown here are averaged for 10 years for June, July, and August.

Climate models must accurately simulate atmospheric convection, which is the rising air motion that causes, in some cases, clouds. Of particular note is a semi-permanent belt of intense convection near the Equator called the ITCZ, or Intertropical Convergence Zone, where winds from the north and south collide and cause rising motion to altitudes as high as 18 km. However, climate models commonly suffer from a problem known as the double-ITCZ, whereby they simulate not one—as is seen in nature—but two ITCZs near the Equator. Most climate models simulate convection using an approach whose basis was developed more than 30 years ago.

In a recent study published in Geophysical Research Letters (23 March 2006), scientists sponsored by the Department of Energy's Atmospheric Radiation Measurement (ARM) Program used ARM data to modify this long-standing approach. This led to clear improvements by largely correcting the double-ITCZ problem. It also corrected some other significant biases shared by a wide range of prominent climate models.

The key to the improvement came from analysis of ARM data from ARM Climate Research Facility sites in the tropics and the Southern Great Plains site in Oklahoma. This analysis showed that—contrary to expectations—convection was not as closely related to fluctuations in atmospheric instability resulting from the daily changes of surface heating and other processes occurring close to the ground as it was to variations in the broad state of the upper atmosphere. This result was used to modify the approach that had been widely used to simulate convection in climate models for the past 30 years.

The improvement was tested in the newest coupled version of the Community Climate System Model from the National Center for Atmospheric Research, which simulates global climate via four primary components that treat the atmosphere, ocean, land-surface, and sea ice. The correction improves the model's ability to compute several important climate features. The most notable improvement is in model-simulated tropical convection, as seen in comparisons between observations and model-simulated rain at the surface (see figure 1). This largely corrects two major biases in the pattern of the simulated rainfall (shared by many prominent climate models) that were in gross disagreement with the observations: (1) eliminating the extra, erroneous ITCZ south of the Equator, and (2) correcting the deficit of rainfall at the Equator. Additionally, the improvement has a profound impact on correcting previous biases in the simulations of sea surface temperatures within the region.