Radiative Kernels

A one-stop shop for radiative kernels from across the climate science community.

Radiative kernels, widely used for diagnosing climate feedbacks and forcing, have been developed from multiple climate models, reanalysis datasets and, most recently, observations. With an aim to improve accessibility of these tools and allow for assessment of the radiative kernel technique itself, we have assembled a collection of existing radiative kernel sets from across the climate science community.


The following link includes a collection of radiative kernel sets for download and a Python-based toolkit for using them*:


University of Miami Radiative Kernel Portal


*No signup required. For consistency, all downloadable radiative kernels have been processed to follow the naming and sign conventions described in the README_KernelFormat file. These may differ from the original radiative kernels available directly from the developers or from versions posted elsewhere.


Radiative Kernels Currently Available for Download

Source TOA Surface Reference Notes

Soden et al. (2008)



Pendergrass et al. (2018)


CloudSat Kramer et al. (2019), in-press Derived from the CloudSat Fluxes and Heating Rates data product.


Radiative Kernel Toolkit

The Radiative Kernel Portal linked above includes a collection of Python code under the "Toolkit" directory that can be used to compute kernel-derived instantaneous radiative forcing and radiative flux changes due to changes in temperature, water vapor, surface albedo and clouds.  These radiative flux changes can be used to compute various metrics relevant to the user and their experimental setup, including traditional climate feedbacks and rapid adjustments.


Angie Pendergrass has also developed a helpful tutorial for calculating radiative feedbacks with the CESM-CAM5 radiative kernels:  https://github.com/apendergrass/cam5-kernels


Radiative Kernels: Background

Fluctuations in Earth’s radiative energy balances are caused by changes in certain surface and atmospheric properties, including surface albedo, temperature, water vapor and clouds. Their contribution to total radiative changes can be diagnosed using radiative kernels, which represent the direct radiative response to a small perturbation in a radiative-relevant climate variable.

Traditionally, radiative kernels have been used to quantify individual climate feedbacks on the top-of-atmosphere (TOA) energy budget. Widespread application to ensembles of climate models has led to many insights regarding sources of uncertainty in climate sensitivity. In recent years, the breadth of radiative kernel applications has grown. For example, radiative kernels defined at the surface (SFC) or in the atmospheric column (TOA - SFC) have been used to diagnose radiative constraints on the global hydrological cycle.

The popularity of the radiative kernel technique stems from its computational efficiency. Once pre-computed with offline radiative transfer calculations, radiative kernels can easily be applied to any model or observational dataset, as long as the necessary variables are available. The radiative kernel technique is not without limitations, however. For example, radiative kernels are subject to radiative transfer error and biases in the model base states typically used to generate them.  This remains a subject of ongoing research. Many of the references in the table above include relevant analysis and discussion.