Atmospheric Data Assimilation: GEOS-5 Adjoint Tools

Traditionally, the impacts of observations on numerical weather forecasts have been measured mainly through observing system experiments (OSEs), in which observations are removed from (or added to) a data assimilation system and the resulting forecasts compared against ones produced using a control set. OSEs have been used successfully to provide a gross measure of the impact of observations on forecasts at various lead times but, because of their expense (multiple executions of the data assimilation system are required), usually involve a relatively small number of independent experiments, each considering relatively large subsets of observations. With this approach it is not practical to examine the impacts of, for example, individual channels on a single satellite instrument, let alone all observations assimilated in a single analysis cycle.

In recent years, new approaches have been developed based on adjoint sensitivities with respect to observations and related methods that can provide more details about the impact of observations, which may be useful to identify problems with an observation or the way it is assimilated. The adjoint technique has proven to be a flexible and computationally efficient way to estimate the impacts of all assimilated observations on a selected measure of short-range forecast error. Although subject to assumptions and limitations inherent in the use of adjoint models, the technique efficiently estimates the impacts of all observations simultaneously, and produces results that can be easily aggregated by data type, location, channel, etc. The technique has gained popularity as a an alternative or complement to traditional OSEs, and is currently used at several forecast centers for experimentation or routine monitoring of the observing system.

Adjoint tools have been used with the GEOS-5 ADAS and the GEOS-5 forecast model in several investigations of the observing system. Adjoint impacts in an OSE framework have highlighted the interdependence of different observations and help interpret OSE results. The adjoint tools have also provided a powerful means to calibrate simulated observations for the GMAO.s OSSE system.

Publications using the GEOS-5 adjoint tools are listed under the publications on the Atmospheric assimilation page.

« Back to Atmospheric Data Assimilation Page