APPLICATION OF ARTIFICIAL NEURAL NETWORK IN THE RESTORATION OF TRITIUM CONCENTRATION IN PRECIPITATION

The artificial neural networks are able to distinguish the complex nonlinear relations between the input/output data. Based on such characteristics, this article selects IAEA/WMO observation data of tritium concentration in atmospheric precipitation from 70 gauging stations in Northern Hemisphere (latitude 22-74°) to establish the restoration model for the annual average concentration of tritium in atmospheric precipitation. With comparison, it is concluded that, the tritium concentration restored by the artificial neural networks can objectively reflect its true value, which provides a new thought for the datum-free areas to restore the tritium concentration in atmospheric precipitation from 1953.