Attribution study on spatially and temporally averaged infrared spectra
Abstract
Direct attribution of changes in spatiotemporal averaged infrared spectra to changes in different climate variables is an emerging approach to climate change detection based on satellite observations. In this study, synthetic AIRS nadir-view longwave spectra as well as radiative kernels for various climate variables are computed and averaged within each 16-day and 10x10 grid box. A requirement to determine climate variable change from mean spectrum changes from two periods is linearity of the spectral radiance change. We test how well the spectra difference can be expressed as the sum of radiative kernels multiplied by climate variable differences (kernel-derived spectra). Results show that based on averaged kernels, nonlinearity errors are not negligible in window region and water continuum absorption band even for clear sky conditions. For all sky conditions, nonlinearity errors are significant throughout the whole spectrum. Relative RMS (root-mean-square) errors of kernel-derived spectra compared to the spectra difference are larger than 30% for 82% of 10° by 10° grid boxes. To mitigate this problem for all-sky conditions, new radiative kernels for cloud properties and complementary kernels to handle the nonlinearity are formulated. Based on the new set of radiative kernels, RMS errors in kernel-derived spectra are reduced to within 4% for all grid boxes. When the new set of kernels is used for the retrieval of surface temperature, temperature and humidity profiles, the retrieval error of climate variable change is significantly reduced. As a consequence, when changes from two consecutive 16-day periods are retrieved in 2009-2010, the temporal correlation between retrieved and true changes is significantly increased. For surface temperature, the correlation coefficient between retrieved and actual changes is larger than 0.9 for 97% of grid boxes. For tropospheric temperature above 850hPa and humidity above 800hPa, correlations are larger than 0.9 for 91% of grid boxes. In boundary layer, the correlation coefficient of temperature is larger than 0.8 except for tropics and Antarctic. The correlation coefficient of humidity changes is poor but reasonable results still can be obtained.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.A13F2513P
- Keywords:
-
- 0305 Aerosols and particles;
- ATMOSPHERIC COMPOSITION AND STRUCTUREDE: 0321 Cloud/radiation interaction;
- ATMOSPHERIC COMPOSITION AND STRUCTUREDE: 0394 Instruments and techniques;
- ATMOSPHERIC COMPOSITION AND STRUCTUREDE: 3305 Climate change and variability;
- ATMOSPHERIC PROCESSES