全球变化知识资源中心

The double Intertropical Convergence Zone bias remains a persistent problem in coupled general circulation model simulations. Due to the strong sea surface temperature (SST)-convection relationship in the tropics, precipitation biases are sensitive to background SST. Using historical simulations of 24 coupled general circulation models and an atmospheric general circulation model, we show that cold equatorial SST biases at least exacerbate double Intertropical Convergence Zone biases in the Pacific. A linear regression model is used to demonstrate that improved predictability of precipitation trends is possible with such model-dependent information as mean-state SST biases accompanying projected SST trends. These results provide a better understanding of the root of the double Intertropical Convergence Zone bias and a possible path to reduced uncertainty in future tropical precipitation trends.

Plain Language Summary Dozens of complex computer models of the climate system (accounting for both atmosphere and ocean) are used for climate change predictions over the coming decades as anthropogenic emissions of greenhouse gases persist. These computer models are essential yet imperfect. Two well-known mismatches (i.e., biases) in models that have persisted for many years include (1) too cold sea surface temperature (SST) along the equatorial Pacific Ocean and (2) excessive precipitation south of the equator, which appears as a double-peaked precipitation pattern known as the double Intertropical Convergence Zone (ITCZ). These are commonly referred to as the cold tongue bias and double ITCZ bias, respectively. Our analysis confirms that they are closely related in models; the worse the cold tongue bias is, the more the ITCZ is split into two. We also found that predicted trends in tropical rainfall depend on these biases; models with the least severe SST bias will improve relative to today's climate as SST warms, whereas the double ITCZ in models with the most severe biases will remain the same even as equatorial SSTs warm. Finally, we demonstrate that rainfall predictions can be improved if information about biases is accounted for.