To predict long term behavior of the climate, scientists study a variety of computer simulations and base themselves on the most common patterns and results. Researcher Svetlana Dubinkina from the Centrum Wiskunde & Informatica (CWI) in Amsterdam, shows that different numerical methods (e.g. methods that are used to simulate flows in ocean and atmosphere on the computer) provide different statistical outcomes. For example to calculate the climatic mean wind field, one choice of a numerical method predicts a weak prevailing wind and another predicts the prevailing state to be wind-still.
Dubinkina studied the statistical accuracy of different numerical methods. Her research shows that the statistics of a simplified model of atmospheric flow are biased by the numerical methods used in the simulation. On May 28 Dubinkina received her PhD degree at the University of Amsterdam for her thesis ‘Statistical Mechanics and Numerical Modeling of Geophysical Fluid Dynamics’. The results of her research are of interest for climate scientists and meteorological institutes.
Dubinkina: “To predict likely behavior of the climate, e.g. the measurement of the average wind speed or the average temperature in winter, one usually looks at the statistics of a model. My research shows that when it comes to long term solution behavior, the statistical measures are highly dependent on the properties of the numerical method that is used. In particular, the ability of those methods to incorporate physical laws determining the atmospheric behavior.”
In her thesis she focuses on early numerical methods by A. Arakawa as well as the very recent ‘Hamiltonian Particle-Mesh Method’, applied to a model that was used for the first weather forecasts by computer. According to Dubinkina, in climate studies the right interpretation of the statistics is only possible when restrictions and properties of method that is used are taken into account. “More research is needed to determine if the results obtained for this simple model can be generalized to real climate models,” she says.