Comparison of surface variables from ERA and NCEP reanalysis with station data over eastern China

This study compared precipitation, mean air temperature (MAT) and mean sea level pressure (MSLP) from two widely used reanalysis datasets (ERA-40 and NCEP) with those from observed stations across eastern China. The evaluation was based on a comparison of both temporal and spatial variability and included several assessment criteria such as the mean values, normalized root mean square error, Mann–Kendall test, empirical orthogonal functions (EOFs) and probability density functions. The results showed that both the ERA-40 and NCEP datasets could capture temporal and spatial variability of the observed precipitation, MAT and MSLP over eastern China. The results showed that the two reanalysis datasets performed better for MAT and MSLP than for precipitation. Overall, the two reanalysis datasets revealed reasonable agreement with observations according to the evaluation. ERA-40 was better at capturing the temporal and spatial distributions for these three variables than NCEP, especially for MAT and MSLP. NCEP tended to overestimate the annual precipitation for both mean and extreme values, while ERA-40 tended to underestimate it, particularly for extreme values. The two reanalysis datasets performed better in the east and northeast regions of the study area than in other regions for capturing the temporal variability of MAT and MSLP. ERA-40 was poor at capturing the temporal variability of precipitation in northeastern China. According to the trend analysis, the two reanalysis datasets showed lower trends for MAT and precipitation and higher trends for MSLP. Both ERA-40 and NCEP had larger explained variances for the first two EOFs than the observed precipitation. This implies that both reanalysis datasets tend to simulate a more uniform spatial distribution for precipitation in the study area.