Simulation of Atmospheric Variability

Abstract

A spectral atmospheric circulation model is time-integrated for approximately 18 years. The model has a global computational domain and realistic geography and topography. The model undergoes an annual cycle as daily values of seasonally varying insolation and sea surface temperature are prescribed without any interannual variation. It has a relatively low computational resolution with 15 spectral components retained in both zonal and meridional directions. Analysis of the results from the last 15 years of the time integration indicates that, in middle and high latitudes, the model approximately reproduces the observed geographical distribution of the variability (i.e., standard deviation) of daily, monthly and yearly mean surface pressure and temperature. In the tropics, the model tends to underestimate the variability of surface pressure, particularly at longer time scales. This result suggests the importance of processes with long time scales such as ocean?atmosphere interaction, in maintaining the variability of the atmosphere in low latitudes. It is shown that global mean values of standard deviation of daily, 5-daily, 10-daily, monthly, seasonal and annual mean surface pressure of the model atmosphere may be approximately fitted by a corresponding set of standard deviations of a red noise time series with a decay time scale of slightly longer than four days. However, it appears that the temporal variation of surface pressure also includes minor contributions from disturbances with much loner decay time scales. In general, the model tends to underestimate the persistence (or decay time scale) of atmospheric disturbances. However, it reproduces some of the features of the observed geographical distribution of decay time scale of the surface pressure fluctuations in middle and high latitudes. The observed standard deviation of annual, hemispheric mean surface air temperature also is compared with model results. Although a clearcut evaluation of model performance is somewhat hampered by observational uncertainty, it appears that the model's value amounts to a substantial fraction of the corresponding standard deviation derived from observational studies.