Assessing the role of observational errors in data assimilation: experiments with a global data assimilation system
Abstract
An approach to experimental investigation of the role of observational errors in data assimilation is proposed. This approach is similar to the widely used OSE and OSSE methodologies. The difference of our method from the above methodologies is that we add simulated pseudo-random errors to the real observations, then run assimilation cycle, assess the assimilation error with respect to the unperturbed observations, and plot an assimilation score against the root-mean-squared magnitude of the added errors. Finally, we extrapolate the resulting curve to zero total (real plus simulated) observational error. In this way, one can obtain an estimate of the potential benefit of improving the accuracy of observations in operational data assimilation. The assimilation system used is described (optimum interpolation is utilised on the analysis step and a 28-level semi-lagrangian NWP model is used to produce the "background" fields in the 6-hour assimilation suite). Results of numerical experiments with perturbed radiosonde observations are presented.
- Publication:
-
EGS - AGU - EUG Joint Assembly
- Pub Date:
- April 2003
- Bibcode:
- 2003EAEJA......277Z