The article begins by going into the history of the use of adjoints in meteorology. Adjoints are considered by many meteorologists to be powerful modeling tools. Sensitivities are a concern of many research meteorologists, and questions surrounding them require estimations of how synoptic features will change (in the modeling) if perturbations are made.
In sensitivity analysis, many methods are at work, but the basic one (as stated above) is that people look back at how a perturbation affected the forecast of a certain disturbance. They can use adjoint models as a way to analyze sensitivity. Adjoint operators can be used for sets of equations that we do not normally call models. Tangent linear models are used to go backwards in time; instead of the standard way people would think which is moving forward in time. The article then goes on to give several examples of adjoint models being put to use. Sensitivity analysis can be a very helpful thing it can be used to figure out forecast model error and input and output variations, among other things. The main limitation is that the adjoint model must be used with the linearization that is valid, otherwise it will not work. Derivation of the correct linearization is complicated, but the future is bright for adjoint models. QUESTIONS:
Where can one access adjoint model findings?
How can adjoint models help me as an individual forecaster?
Errico, Ronald M. 1997: What is an adjoint model? Bulletin of the American Meteorological Society Vol. 78, 11, 2577-2591.