Band-by-band longwave flux and cloud radiative forcing at the top of atmosphere: observation vs. simulation

We present a study of the top-of-atmosphere flux and cloud radiative forcing(CRF) over each individual longwave absorption band (hereafter termed as band-by-band longwave fluxes and CRFs) as derived from collocated AIRS and CERES observations and as simulated from the GFDL GCM. Such band-by-band radiation budgets and CRFs are what climate models directly compute. Thus evaluating and tuning GCMs with such quantities brings no additional efforts for modeling community, and is more rigorous than broadband flux/CRF because it avoids the compensating biases from each band, a feature commonly seen in the comparisons of broadband flux/CRF. Moreover, while the longwave broadband CRF is equally sensitive to cloud fraction and cloud top height, the fractional contribution of each band to the longwave broadband CRF is sensitive to cloud top height but largely insensitive to cloud fraction, which makes it a meaningful metric to explore.

AIRS and CERES on Aqua provides first opportunity of collocated and concurrent global observations of both broadband longwave radiance and spectrally-resolved radiance over the majority of longwave spectrum. We derive band-by-band logwave TOA fluxes and CRFs over the tropical oceans from the AIRS spectra with the help of well-established CERES broadband measurements. Based on the predefined scene types in the CERES-SSF dataset, spectrally-dependent ADMs are developed and used to estimate the spectral flux for each AIRS channel. A multivariate linear prediction scheme is then used to estimate spectral fluxes at frequencies not covered by the AIRS instrument. As a result, spectral fluxes over 10cm^-1 interval are derived from the entire longwave spectrum. The whole algorithm is validated using synthetic spectra as well as the CERES OLR measurements. The differences between OLR computed by this algorithm and collocated CERES OLR are 0.67±1.52 Wm^-2 for clear sky and 2.15±5.51Wm^-2 for cloudy sky. The algorithm behaves consistently well over times and over difference cloudy scene types, indicating the robustness of the algorithm.

Then, using the GFDL AM2 model as a case study, we illustrate the merit of band-by-band fluxes and CRFs derived from this algorithm in model evaluation. By comparing the observed spectral fluxes and simulated ones for the year of 2004, compensating errors in the simulated OLR from different absorption bands are clearly revealed. The AM2 tropical annual-mean band-by-band CRFs generally agree with the observed counterparts, but some systematic biases in the window bands and over the marine-stratus regions can be clearly identified. Furthermore, the interannual variations of such band-by-band CRFs are studied by EOF analysis. The leading mode is clearly associated with ENSO. When contribution of each band to the broadband CRFs is examined, the EOF spatial pattern is different from band to band, revealing the change of cloud top height in response to the ENSO cycle.

Figure caption: Upper panels: The annual-mean longwave broadband CRF from the AIRS&CERES collocated observations (left side) and from the AM2 simulation (right side). Middle panels: The fractional contribution of water vapor rotational band and vibrational v2 band to the annual-mean longwave CRF. Lower panels: Same as the middle ones except for 1070-1200cm^-1, a window band.