The evolution of ERA-40 surface temperatures and total ozone compared to observed Swiss time series

The aim of this paper is to compare long term changes derived from ERA-40 reanalysis with the correspondin, changes found in carefully homogenized observational data. Focus is given on two climate key parameter, daily surface temperature (T2) and total ozone in the Alpine region for the period 1961-2000. Overall the two temperature data sets agree well. The reanalysed temperature trend (0.3 degrees C/decade) is similar to the trend in raw data but underestimates the trend derived from homogenized data by 0.1 degrees C/decade. Differences between ERA-40 and the observations decrease with time. The bias is reduced from +/- 2 degrees in the 1960s to +/- 1 degrees in the 1990s. Differences show a distinct annual cycle. On average ERA-40 overestimates T2 in summer by 1.0 degrees C and underestimates it by 0.4 degrees C in winter. A simple downscaling procedure is applied to compare the temperature values of single Swiss station series with ERA-40 data. The technique makes use of a topographic lapse rate derived from observations. It is shown that the lapse rate has a distinct seasonal cycle in the Alpine region. The use of this seasonal varying lapse rate reduces considerably the seasonal bias found between ERA-40 and the observed series. In contrast to the temperature data the two ozone data sets show strong differences between time periods satellite data has been assimilated or not. In the episodes where ozone measurements from satellites are used, basically after 1979, the total ozone values agree well with ERA-40. The reanalysis data slightly overestimates the observations by 10 DU (Dobson Units). Before 1973 the error is a little bit larger with 12 DU, but the lack of satellite data assimilation in ERA-40 results in an enhanced seasonal bias with maximum errors in winter (about 20 DU). The strongest differences are found in the 1970s, when radiance from satellite data was assimilated, but ozone was not. There is a mean bias of 38 DU with maximum errors in winter of about 70 DU.