A Comparison Of SSM/I-Derived Global Marine Surface Specific Humidity Datasets

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015
John Prytherch, University of Leeds, Leeds, United Kingdom; and E. C. Kent, S. Fangohr, and D. I. Berry

Handout (14.7 MB)

Observations of near-surface humidity over the oceans are needed for climate and air-sea interaction applications. Observations from ships have been used to generate fields of near surface specific humidity (qa) in datasets based on observations archived in the International Comprehensive Ocean-Atmosphere Data Set (ICOADS). Such in situ-based humidity datasets have been produced by the University of Wisconsin at Milwaukee, by the UK National Oceanography Centre (NOC), and by the Centre for Ocean-Atmosphere Prediction Studies at the Florida State University. These datasets provide long time series estimates of qa, but are compromised by poor sampling in some regions, especially in the Southern Hemisphere. Using satellite observations to provide wider coverage for qa would therefore be extremely valuable.

Satellite-based microwave sensors, such as the Special Sensor Microwave Imager (SSM/I), have provided a means to retrieve qa since the 1980s. Seven publically available satellite measurement-derived monthly mean humidity datasets are compared with one another and with in situ qa from the NOC Surface Flux and Meteorological dataset. The datasets are shown to be markedly different in terms of their large-scale means, and their smaller-scale spatial and temporal structure Differences show a range of yearly, global mean qa of ~ 1 g kg-1. Comparison of the datasets derived using the same satellite measurements of brightness temperature reveals differences in qa that depend on the source of satellite data; the processing and quality control applied to the data; and the algorithm used to derive qa from the satellite measurements of brightness temperature. Regional differences between satellite-derived qa due to the choice of input data, quality control, and retrieval algorithm can all exceed the accuracy requirements for surface flux calculation of ~ 0.3 g kg-1 and can be several g kg-1 in monthly means for some periods and regions.