On the Multi-scale Variability of High-frequency Surface Air Temperature

Abstract:

We demonstrate that the first four statistical moments of sub-daily surface air temperature (SAT) anomalies exhibit large spatial patterns, globally, which differ from moment-to-moment and that many regions have statistically significant trends in moments from 1950-2010; these results imply that high-frequency SAT anomaly distributions are nearly identically distributed over very large spatial scales and that these distributions are undergoing characteristic changes in shape due to either decadal variability or climate change. Further, we examine the spatial scaling structure of higher-order and non-linear spatial correlations up to fourth-order which determine the variability distributions of SAT at larger spatial scales. Higher-order moment statistics suggest that SAT scales as an approximately locally homogeneous and isotropic quasi-Gaussian random field whose higher-order moments can be determined by functions of pair correlations, which in turn are related to regionally varying decorrelation length scales. These results have implications for the study of multi-scale atmospheric variability, extremes, and climate change involving geographically smooth variables and helps to define the theory which underlies the success of statistical downscaling techniques.