Observing system simulation experiments (OSSEs) have been traditionally used to measure the anticipated impact of a new set of measurements on weather prediction. Forecast OSSEs are useful, in that they measure the effectiveness of a set of measurements in the context of the current global observing system. However, forecast OSSEs also have a number of key limitations. First, they require simulation of all current measurements, along with calibration of their errors. Second, they rely on the availability of a forecast and data assimilation system that is capable of ingesting the new measurements. If measurement uncertainties are not properly calibrated, the measure of impact of a new observing system will be incorrect. If a data assimilation system is not capable of assimilating a new type of measurement, a forecast OSSE is not possible. In fact, there is an inherent paradox: measurements are needed for processes that are not well understood, while poorly understood processes are generally not realistically simulated by numerical models.
In this presentation, we propose a hierarchy of OSSEs that range from simple to complex, and include:
- Experiments that explore satellite spatial and temporal sampling
- Quantification of retrieval (and forward model) uncertainty
- Assessment of mission science goals
- Observation impact on numerical weather prediction
We use as an example the measurement of the three-dimensional distribution of atmospheric winds from space, which has been highlighted as a key observational need during the coming decade.