This thesis has initiated the fusion of neural network (NN) techniques
and dynamical models. With a 6-layer dynamical ocean model coupled to an
NN atmosphere-- the first hybrid coupled model (HCM) with a nonlinear
empirical atmospheric component has been developed and named the
NHCM. The NHCM showed that the nonlinear atmosphere could have
advantages over a linear atmosphere in modelling ENSO (El Nino-Southern
Oscillation) variability and in predicting ENSO events.

For better forecast skills, different types of data-- surface wind stress,
upper ocean heat content anomaly (HCA), sea surface temperature (SST) and
sea surface height anomaly-- have been assimilated into the coupled model
using a 3D-Var data assimilation scheme. The results show that assimilating
HCA yields overall the greatest improvement in the forecast correlation
skills. This thesis has lead to an operational ENSO prediction system, which
has been used to issue routine forecasts of the tropical Pacific sea
surface temperatures.

While the full 4-D var HCM is beyond the scope of this thesis, a
neural-dynamical hybrid approach under 4-D var has been developed to
study the simple Lorenz system.

For more info., please visit http://www.cims.nyu.edu/~ytang