资源环境科技发展态势分析平台

Understanding future changes in climate variability, which can impact human activities, is a current research priority. It is often assumed that a key part of this effort involves improving the spatial resolution of climate models; however, few previous studies comprehensively evaluate the effects of model resolution on variability. In this study, we systematically examine the sensitivity of temperature variability to horizontal atmospheric resolution in a single model (CCSM3, the Community Climate System Model 3) at three different resolutions (T85, T42, and T31), using spectral analysis to describe the frequency dependence of differences. We find that in these runs, increased model resolution is associated with reduced temperature variability at all but the highest frequencies (2-5 day periods), though with strong regional differences. (In the tropics, where temperature fluctuations are smallest, increased resolution is associated with increased variability.) At all resolutions, temperature fluctuations in CCSM3 are highly spatially correlated, implying that the changes in variability with model resolution are driven by alterations in large-scale phenomena. Because CCSM3 generally overestimates temperature variability relative to reanalysis output, the reductions in variability associated with increased resolution tend to improve model fidelity. However, the resolution-related variability differences are relatively uniform with frequency, whereas the sign of model bias changes at interannual frequencies. This discrepancy raises questions about the mechanisms underlying the improvement at subannual frequencies. The consistent response across frequencies also implies that the atmosphere plays a significant role in interannual variability.