Spectral radiative kernel technique and the spectral dimension of radiative feedback

In recent years the broadband radiative kernel technique has been established as an efficient way to compute broadband radiative feedbacks from a variety of GCM (general circulation model) output. The spectrum is an intrinsic dimension of radiative feedback since it is the integrand of broadband radiative flux. In this study, we address two questions: (1) Can we compute the spectral radiative feedback in a way as efficient as the broadband feedback is computed using radiative kernel technique? (2) What unique insights can we gain from the spectral dimension of radiative feedback, which cannot be obtained from studying broadband radiative feedback alone? To answer question (1), we develop a set of longwave spectral radiative kernel (SRK) at 1cm-1 spectral resolution based on the GFDL model simulation and a fast spectral flux simulator based on the PCRTM model. The technique approach of SRK is similar to that of broadband radiative kernel, but with the spectral dimension explicitly computed. We validate such SRK technique against more rigorous partial radiative perturbation (PRP) methods and, for both clear-sky and all-sky feedbacks, the SRK technique can agree with PRP methods within 10% for all frequencies in the longwave. We also use the SRK method to compute the LW broadband feedbacks of GCMs that participated in CMIP3 and CMIP5 and we show the results are all consistent with published studies using broadband radiative kernel technique. To answer question (2), we use the validated SRK to compute longwave spectral radiative feedbacks using the monthly-mean output of CMIP3 and CMIP5 archives. It is found that, although lapse rate and water vapor broadband feedbacks are always anti-correlated with each other, their spectral shapes are indeed different considerably. Moreover, at certain spectral bands, the discrepancies in spectral lapse rate feedbacks among different GCMs can be as large as a factor of 8-10, much larger than the discrepancies in lapse rate broadband feedback. This finding has implications for both GCM evaluations and observational strategy for monitoring climate changes.