Method for reconstruction of monthly average data on water transparency for the North-West Black Sea
A.S. Lubkov1, E.N. Voskresenskaya1, A.S. Kukushkin2
1Natural and Technical Systems Institute of the RAS, 28, Lenina St., Sevastopol, 299011, Russia
2Marine Hydrophysical Institute of the RAS, 2, Kapitanskaya St., Sevastopol, 299011, Russia
Keywords: прозрачность, нейронные сети, речной сток, гидрометеорологические условия, Черное море, transparency, neural network, runoff, hydrometeorological conditions, Black Sea
Abstract
To reconstruct the data of hydrophysical parameter observations, a model based on a neural network with the teacher has been suggested. The indices of global climate modes of the ocean-atmosphere system were applied as the model input. The processes of the model teaching and adaptation, which allow one to find the most accurate solution of the problem of modeling, is described. The comparison of modeled monthly average Danube's runoff with data of observation demonstrated their good correspondence. Missed in some years observation data on sea water transparency (depth of white disc visibility) were reconstructed. The proximity of reconstructed and observed depths of white disc visibility was noted. Some features of interannual variability of reconstructed data on sea water transparency caused by both climatic factors during 1950-1962 and changes in the chlorophyll а concentration during 1998-2010 were found.
|
