Seminar: Katarina Kosovelj: A scale-dependent evaluation of the bias in a climate model

Tuesday, 10th January 2017, 15.15
Room F6, 1st floor, FMF, Jadranska 19, Ljubljana

Meteorology Group will host a seminar presented by Katarina Kosovelj. The target audience are masters students but everyone is kindly invited.


Although all climate models are useful, their performance vary. As we do not know the future states of the atmosphere, climate models have to be compared with the past and present climate in order to be validated.
One of the common methods is the evaluation against reanalysis as they are the best estimate of the past climate available. The averaged differences between a climate model and a reanalysis represent the model bias. Spatial patterns of biases are helpful for the understanding of the model deficiencies.
If the model biases are calculated in spectral space, it makes possible their scale-dependent evaluation i.e. biases in each wavenumber. This information complements the evaluation in grid-point space.
In the seminar I will present such evaluation of a general circulation model SPEEDY against ERA 20C reanalysis data for the period 1911-2010. The SPEEDY projection to spectral space was performed with the MODES software which allows us to split the biases into components associated with the quasi-geostrophic and inertia-gravity circulations. Each part can be then evaluated separately, also in grid-point space.
Results show that biases in SPEEDY are large. This is not surprising since it is a simplified research model run on a coarse resolution that lacks complexity of the full-scale reanalysis models.