On statistical dissipation in GCM-climate

This work studies the dissipation that affects the statistical behavior of a climate variable. The hypothesis tested is that the statistical dissipation of large-scale variables in a climate model is significantly influenced by the temporal variations of the modelʹs small scale variables. The test is made with the T21 ECHAM4 atmospheric GCM by varying modelʹs representation of small-scale variables in two different ways. First, the strength of the horizontal diffusion is modified. Secondly, white noise unrelated to the state of the large-scale variables is added to variables with the smallest scales (wave numbers equal to and smaller than 18). It is found that the statistical dissipations of large-scale variables (e.g., vorticity at wave numbers equal to and larger than six) depend on the intensity of small-scale fluctuations, no matter whether they are induced by modifying the horizontal diffusion or by adding noise. The stronger the small-scale fluctuations, the stronger are the dissipations of the largescale variables. This result suggests that the simulation of low-frequency climate variations and the prediction of climate change responses depend on the model representation of small-scale climate components.