Improved multisensor approach to satellite-retrieved near-surface specific humidity observations

A multisensor microwave retrieval of near-surface 10 m specific humidity (Qa) using satellite observations from the advanced microwave sounding unit-A (AMSU-A), the Special Sensor Microwave Temperature Sounder-2, and the Special Sensor Microwave Imager is improved upon in this study. Refinements to the regression formula, training data set, collocation procedure, and height adjustment to 10 m were used to improve retrievals from two different sensor combinations. Independent validation with the International Comprehensive Ocean-Atmosphere Data Set (ICOADS) indicates a lower overall bias of similar to 0.3 g/kg and a smaller root-mean-square difference of similar to 0.4 g/kg than with several previously published single-sensor Qa retrievals. A significant regional Qa wet bias of similar to 3 g/kg in the summer over the North Pacific was found for all satellite retrievals, and a correction was developed using an inversion index defined using sea surface temperature and AMSU-A lower tropospheric temperature observations. An assessment of ICOADS ship and buoy validation data indicated uncertainties related to height adjustments of these in situ observations to be 0.2-0.4 g/kg, while hygrometer differences and solar heating effects had smaller uncertainties of less than 0.05 g/kg. Validation of the updated multisensor retrievals with ICOADS over an 8-year period from 1999 to 2006 revealed a reduced magnitude of the regional biases when compared to previously published retrievals. Regional Qa differences, particularly in the subtropical high regions, are shown to play a significant role in determination of surface latent heat flux.