Abstract
Radiative kernels describe the differential response of radiative fluxes to small perturbations in state variables and are widely used to quantify radiative feedbacks on the climate system. Radiative kernels have traditionally been generated using simulated data from a global climate model, typically sourced from the model's base climate. Consequently, these radiative kernels are subject to model bias from the climatological fields used to produce them. Here, we introduce the first observation‐based temperature, water vapor, and surface albedo radiative kernels, developed from CloudSat's fluxes and heating rates data set, 2B‐FLXHR‐LIDAR, which is supplemented with cloud information from the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO). We compare the radiative kernels to a previously published set generated from the Geophysical Fluid Dynamics Laboratory (GFDL) model and find general agreement in magnitude and structure. However, several key differences illustrate the sensitivity of radiative kernels to the distribution of clouds. The radiative kernels are used to quantify top‐of‐atmosphere and surface cloud feedbacks in an ensemble of global climate models from the Climate Model Intercomparison Project Phase 5, showing that biases in the GFDL low clouds likely cause the GFDL kernel to underestimate longwave surface cloud feedback. Since the CloudSat kernels are free of model bias in the base state, they will be ideal for future analysis of radiative feedbacks and forcing in both models and observations and for evaluating biases in model‐derived radiative kernels.
Key Points
Observation‐based radiative kernels are introduced, providing a method for quantifying feedbacks with reduced model bias
Accounting for the cloud masking of noncloud feedbacks is critical for interpreting the sign of longwave cloud feedback
The distribution of clouds in the polar region contributes to the seasonality of shortwave cloud feedback
