Air temperature and relative humidity have been the main parameters of meteorology study. In the past data could be obtained from in-situ observations, but the observations are local and sparse, especially over ocean....Air temperature and relative humidity have been the main parameters of meteorology study. In the past data could be obtained from in-situ observations, but the observations are local and sparse, especially over ocean. Now we can get them from satellites, yet it is hard to estimate them from sat- ellites directly so far. This paper presents a new method to retrieve monthly averaged sea air temper- ature （SAT） and relative humidity （RH） near sea surface from satellite data with artificial neural networks （ANN）. Compared with the observations in Pacific and Atlantic, the root mean square （RMS） and the correlation between the estimated SAT and the observations are about 0.91 ~C and 0.99, respectively. The RMS and the correlation of RH are about 3.73% and 0.65, respectively. Compared with the multiple regression method, the ANN methodology is more powerful in building nonlinear relations in this research. Thus the global monthly average SAT and RH are retrieved from the fixed ANN network from July 1987 to May 2004. In general the annual average SAT shows the increasing trend in recent 18 years. The abnormality of SAT is decomposed with the empirical or- thogonal function （EOF）. The leading three EOFs could explain 84% of the total variation. EOF1 （76.1%） presents the seasonal change of the SAT abnormality. EOF2 （4.6%） is mainly related with ENSO. EOF3 （3.3%） shows some new interesting phenomena appearing in the three main currents in Pacific, Atlantic and Indian Ocean.展开更多
基金Supported by The National Key Technology R&D Program(No.2013BAD13B01)the National High Technology Research and Development Program of China(No.2001AA633060)
文摘Air temperature and relative humidity have been the main parameters of meteorology study. In the past data could be obtained from in-situ observations, but the observations are local and sparse, especially over ocean. Now we can get them from satellites, yet it is hard to estimate them from sat- ellites directly so far. This paper presents a new method to retrieve monthly averaged sea air temper- ature （SAT） and relative humidity （RH） near sea surface from satellite data with artificial neural networks （ANN）. Compared with the observations in Pacific and Atlantic, the root mean square （RMS） and the correlation between the estimated SAT and the observations are about 0.91 ~C and 0.99, respectively. The RMS and the correlation of RH are about 3.73% and 0.65, respectively. Compared with the multiple regression method, the ANN methodology is more powerful in building nonlinear relations in this research. Thus the global monthly average SAT and RH are retrieved from the fixed ANN network from July 1987 to May 2004. In general the annual average SAT shows the increasing trend in recent 18 years. The abnormality of SAT is decomposed with the empirical or- thogonal function （EOF）. The leading three EOFs could explain 84% of the total variation. EOF1 （76.1%） presents the seasonal change of the SAT abnormality. EOF2 （4.6%） is mainly related with ENSO. EOF3 （3.3%） shows some new interesting phenomena appearing in the three main currents in Pacific, Atlantic and Indian Ocean.
