Subjective versus objective sensitivity estimates: Application to a North African cyclogenesis

An observing system simulation experiment is used to test and compare objective and subjective estimates of sensitivity of a forecast aspect to the initial condition (IC) fields for a case of rapidly developing cyclogenesis over the Western Mediterranean during 19-22 December 1979. The ability of sensitivity estimation methods to provide helpful guidance about where an improvement in the IC can lead to the largest forecast error reduction is particularly important to ascertain in order to guide adaptive observation campaigns. Synthetic soundings from a 15-km reference simulation are added to an initially poor 60-km control simulation over the sensitive areas as determined by the combination of the given sensitivity estimate and a simple analysis error estimate. The ability of each sensitivity estimation method to produce an improved simulation of the cyclone is assessed. Results show that while the sensitivity estimates perform similarly, with no significant differences among them, the subjective method yields the best overall targeting guidance. In contrast, the adjoint estimate provides the least accurate targeting guidance for this particular case and analysis error estimate. This suggests that subjective sensitivity estimation methods are able to compete with or even improve upon the objective estimation method for this case of cyclogenesis over the Western Mediterranean.