气候变化领域集成服务门户

The variability of the surface wind field over Northeastern North America was analysed through a statistical downscaling (SD) approach, using the relationships among the main large-scale and observed wind circulation modes. The large-scale variables were provided by 12 global reanalyses. The observed zonal and meridional wind components come from a database of 525 sites spanning over 1953-2010. A large percentage of the regional variability was explained in terms of three major large- and regional/local-scale coupled circulation patterns, accounting for 55.3% (59.3%) of the large (regional/local) scale variability. The method delivered robust results regardless of the SD model configuration, albeit with sensitivity to the number of retained circulation modes and the large-scale window size, but not to the reanalysis chosen for the large-scale variables. The methodological uncertainty was larger for sites/wind components with larger variability. A parameter configuration chosen for yielding the best possible SD estimations showed high correlation values between these estimations and the observations for the majority of the sites (0.6-0.9, significant at p<0.05), and a realistic wind variance (standard deviation ratios between 0.6 and 1.0), with similar results regardless of the reanalysis. The reanalysis direct wind outputs showed higher correlations than the SD estimates (0.7-0.97, also significant). The skill in reproducing observational variance differed considerably from model to model (ratios between 0.5 and 3). The regional wind climatology was reconstructed back to 1850 with the help of century long reanalyses and two additional SLP gridded datasets allowing to estimate the variability at decadal and multidecadal timescales. Recent trends in the wind components are not unusual in the context of century-long reconstructed variability. Extreme values in both components tend to appear associated with high values in the first two modes of variability.