USE OF NEURAL NETWORKS TO IMPROVE COMPUTATIONAL EFFICIENCY OF ENVIRONMENTAL NUMERICAL MODELS

A new generic approach to improve computational efficiency of certain processes in numerical environmental models is formulated. This approach is based on the neural networks (NN) technique. It can be used to accelerate the calculations and improve the accuracy of the parameterizations of several types of physical processes which generally require computations involving complex mathematical expressions, including differential and integral equations, rules, restrictions and highly nonlinear empirical relations based on physical or statistical models. It is shown that, from a mathematical point of view, such parameterizations can usually be considered as continuous mappings (continuous dependencies between two vectors). It is also shown that NNs are a generic tool for fast and accurate approximation of continuous mappings and, therefore, can be used to replace primary parameterization algorithms. In addition to fast and accurate approximation to the primary parameterization, NN also provides the entire Jacobian for very little computation cost. Two particular applications of the NN approach are presented here: (1) a NN approximation of the UNESCO equation of state of the sea water (density of the seawater) and an inversion of this equation (salinity of the seawater); and (2) a NN approximation for th