Environmental Science and Engineering Seminar
Models used to predict climate behavior must be rigorously vetted against observations, the spectral characteristics of global surface temperature being a critical benchmark in this regard. Temperature spectra are known to obey scaling laws connecting forcings of astronomical origin, from orbital to diurnal scales. Many past studies have raised concerns that this continuum is misrepresented by the current generation of climate models. In this talk I will update this picture with recent data syntheses and model simulations, showing that the observed spectrum of global mean temperature is in fact consistent with models of a wide range of complexity, from yearly to orbital scales. This implies that the current hierarchy of climate models harbors no fundamental deficiency regarding energy transfers between timescales, in spite of regional biases. I will suggest that successful climate predictions at decadal to centennial horizons hinge critically on the accuracy of both initial and boundary conditions, particularly for the ocean state. Strategies to usefully estimate pre-instrumental climate states will be discussed.