The common practice in dynamic downscaling is to nest a higher‐resolution regional climate model (RCM) into a global model that resolves the large‐scale circulation. However, nested RCMs can develop distinct large‐scale features that substantially diverge from those of the driving model. This is especially problematic over regions such as South America (SA), where the climate features strong teleconnection with remote oceans. Here we propose to explicitly resolve the atmospheric processes underlying the teleconnection by expanding the RCM domain to include the influential oceans. Using the coupled RegCM4.3.4‐CLM4.5 model, RCM simulations designed under the new paradigm demonstrate a substantial improvement of model skills over those using the standard CORDEX SA domain. Analysis of the underlying physical mechanisms indicates that the RCM captures the large‐scale dynamics and climate teleconnections substantially better when it includes the influential oceans. The Big Brother experimental protocol is then used to identify sources of uncertainties and skills, and the results suggest that the nesting practice cannot effectively capture the impact of forcings and processes acting outside the RCM domain. This uncertainty introduces substantial systematic bias to RCM simulations yet is not sampled by existing coordinated regional modeling projects (e.g., CORDEX) due to the use of a single domain focusing over land. Explicitly including oceans within the domain substantially reduces the sensitivity of the SA model climate to domain size/location and promises great potential for RCM applicability in studying regional mechanisms and feedback processes of SA's hydroclimate.

中文翻译：

---

明确考虑到远洋在南美区域气候模拟中的作用

动态降尺度的常见做法是将高分辨率的区域气候模型（RCM）嵌套到解决大规模环流的全球模型中。但是，嵌套的RCM可以开发出明显的大型特征，这些特征与驾驶模型的特征大不相同。在诸如南美（SA）之类的地区，这尤其成问题，那里的气候特征是与遥远的海洋之间存在紧密的遥远联系。在这里，我们建议通过扩展RCM域以包括影响性海洋来明确解决远程连接背后的大气过程。使用耦合的RegCM4.3.4-CLM4.5模型，在新范式下设计的RCM仿真表明，与使用标准CORDEX SA域的模型相比，模型技能得到了显着提高。对基本物理机制的分析表明，当RCM包括有影响的海洋时，它可以更好地捕获大规模动态和气候遥相关。然后，使用“大哥大”实验协议来识别不确定性和技能的来源，结果表明，嵌套实践无法有效地捕获RCM域外作用力和过程的影响。这种不确定性给RCM仿真带来了实质性的系统偏差，但由于使用了一个专注于土地的单一域，因此尚未被现有的协调区域建模项目（例如CORDEX）进行抽样。