hide
Free keywords:
El Nino; forcing experiment; GCM; neural network; signal separation; volcanic effect
Abstract:
The statistical separation of volcanic and El Niño effects in the atmospheric general circulation model (GCM) data, is based on comparing the combined forces simulation and the output of the individual forcing experiments. These data are provided by the Hamburg Max-PIanck-Institut GCM version ECHAM2 T21.The GCM experiments concern the Norhern Hemisphere winter. The statistical methods for signal separation are linear regression and neural networks. It is found, that in most parts of the atmosphere the effect of the simultaneously applied forces equals the linear sum of the individual effects. This holds for the Southern Hemisphere, the tropics and the Pacific area of the Northern Hemisphere. The linear behaviour is found at all levels of the model atmosphere. The nonlinearities, detected by the neural network, are restricted to the Atlantic European area and the polar region. The neural network strongly improves signal separation especially for the geopotential height (500 hPa). A physical mechanism is presented, which relates the nonlinearities in signal separation to the interaction of the volcanically forced stratospheric westerlies and the El Niño forced subtropical jet.