Reconstruction of Northern Hemisphere 500?hPa geopotential heights back to the late 19th century

Summary In this study the authors have developed a statistical method and have reconstructed Northern Hemisphere 500 hPa heights back to the late 19th century using one temperature and three sea level pressure (SLP) data sets. First, the relationship between ERA40 500 hPa heights and surface temperature and SLP was screened using stepwise multiple regression based on the calibration period of 1958–2002 (1998/2000 according to the availability of SLP data). All selected predictors (temperature and SLP) were significant and their variance contribution was greater than 1%. On average, there were 8.1 variables retained in the final regression equations. Second, the regression equations were applied to compute the 500 hPa height through to the late 19th century for the whole Northern Hemisphere. As the SLP and temperature coverage improved over time, the number of predictors decreased by about 1 in the most recent periods, and the root mean squared error decreased by about 0.8 m. A leave-one-out cross-validation method was used to test the skill and stability of the regression models. The reduction of error during the cross-validation period of 1958–1997 varied from 0.33 to 0.56, depending on the SLP data. Reconstructions were also checked using NCEP/NCAR 500 hPa heights from January 1949 to December 1957, and compared with the historical reconstruction over Europe. Reconstructions show high consistency with these independent data sets. Generally, the reconstruction provides a valuable opportunity to analyze, as well as to validate climate simulations of the variability in free atmosphere circulations over the past one hundred years.