The ability of a climate model to capture realistic interannual variability is an important measure of its performance. Here, we use the standard deviation of yearly annual and seasonal (DJF, JJA) means simulated for the 20th Century to indicate the magnitude of the model's interannual variability. For comparison, NCEP-NCAR Reanalysis Project 1 data for the years 1950-1999 are also presented.
The standard deviation reveals only the magnitude of the variability, and tells nothing about the time scale of that variablity. In the case of CCSM3, for example, we know from other studies that the time scale of ENSO-like variability is too short.
Results are shown for temperature and precipitation in the bar charts below. Raw data are provided in the corresponding tables. Values for "Annual" describe the interannual standard deviation in annual average temperature or precipitation, while summer and winter are for the interannual variation in seasonal data. Standard deviation in precipitation was computed using the total precipitation in cm for a season or year.
Using the NCEP reanalyses as a baseline, we find that all models
show a reasonable magnitude of interannual variability. Only HADCM3
fails to capture the relative magnitudes of the annual, djf, and jja
variability. Interestingly, for precipitation, annual variablity is
greater than for either season. This result could indicate that the
equinoxes contribute significantly or that summer and winter are
strongly correlated.
|
|