资源环境科技发展态势分析平台

Most sounding channels sensitive to atmosphere layers close to Earth's surface are also sensitive to Earth's surface properties. Biases and uncertainties in Earth's surface emissivity and skin temperature may degrade the values of these observations being assimilated into weather prediction models. A method that combines several individual channels into a synthesized channel is proposed here to reduce such uncertainties. The effectiveness of such channel-synthesizing method is first demonstrated through perfect model experiments, where brightness temperatures are simulated and compared before and after noises added to surface emissivity and skin temperature. Real-case experiments that compare simulated brightness temperature and satellite observations further show that the synthesized channel can effectively reduce the mean bias of simulated brightness temperature from 1 to 3K for individual GOES-R channels to near zero for the synthesized channel, suggesting great potential of the approach for more effective assimilation of surface-sensitive sounding channels.

Plain Language Summary Satellite observation is one of the most impactful sources of observations assimilated into weather prediction models. Satellite channels sensitive to atmosphere layers close to Earth's surface are usually sensitive to the Earth's surface properties. Due to the large uncertainty in Earth's surface emissivity and skin temperature, these satellite channels are not well utilized. Here we propose a novel method that combines several individual channels into a synthesized channel to reduce the sensitivity to Earth's surface properties by calculating coefficients for each channel so that influences from Earth's surface properties cancel. For example, if the errors in calculated brightness temperature of three channels due to surface emissivity error are 5, 2, and 4K, respectively, and errors due to skin temperature error are 4, 2, and 3.5K, respectively, a synthesized channel equals 2*CH1+3*CH2-4*CH3 will be used so that the errors cancel. Our real-case experiments using GOES-R observations show that this channel-synthesizing method can effectively reduce mean bias of simulated brightness temperature from 1 to 3K for individual channels to near zero for the synthesized channel, suggesting great potential of the approach for more effective assimilation of surface-sensitive sounding channels.