Efficient methods to assimilate remotely sensed data based on information content

Two basic approaches have evolved to utilize measurements of radiance (i.e. thermal or scattered solar radiation) by satellite-borne instruments in data assimilation systems: radiances (raw or cloud-corrected) may be assimilated directly, or they may be pre-processed to retrieve geophysical parameters for subsequent assimilation. The retrieval process is often ill-posed, and therefore requires the use of prior information to constrain the solution. For example, temperature and humidity profiles retrieved using radiances from nadir-viewing infrared and microwave sounders often incorporate prior information in the form of climatology or forecasts. The use of prior information presents difficulties when assimilating retrievals. Here we present methods to remove prior information from retrievals in order to achieve a more consistent assimilation of the data. In addition, these methods can be used as a data compression device, which can reduce the amount of computation required by some analysis systems compared with radiance assimilation. The methods are implemented and compared in a one-dimensional assimilation system using simulated data from current and future infrared-temperature profiling instruments.