How Does one Build and Adjoint for ROMS/TOMS?

Hernan G. Arango
IMCS, Rutgers University

An overview of the current ROMS/TOMS framework will be presented with particular emphasis to its associated tangent linear (TLM), representer (RPM), and adjoint models (ADM). These new models are used, in conjunction with the nonlinear model (NLM), for stability and sensitivity analysis, variational data assimilation, and ensemble prediction. The complexity of the ROMS/TOMS has increased tremendously and any future change to the basic algorithm has to be propagated to four different but related models. Therefore, it is important to know how all these new models are build. The building of these algorithms is shown in a simple example with animations. Several technical issues regarding the efficiency, parallelization, and testing will be addressed.