All Data are Useful, but not All Data are Used! What´S Going on Here?

Issues limiting the use of modern data streams are described and possible approaches are outlined. Data assimilation system constraints and limitations are listed; a taxonomy of the various errors involved is given; and the difficulty of showing positive impacts for new or currently unused data is discussed. Improvements may come from creative, parsimonious representations of the principal errors; adaptively selecting or pre-processing the data; and questioning the basic assumptions made in current operational data assimilation systems. Examples given from the perspective of satellite scatterometry and IR retrievals illustrate problems of general interest.