Using Native Error Diagnostics to Identify Sources of Aerosol Radiative Forcing Error in GCMs and Understand Their Implications

Thursday, December 13, 2018 - 15:10
Add to Calendar

The range of diagnosed aerosol radiative forcing has remained at a persistently large value of 2 W/m2 across at least the last two Coupled Model Intercomparison Project (CMIP) ensembles. To break the apparent deadlock in aerosol radiative forcing uncertainty, we seek to address drivers of this uncertainty through analysis of the Radiative Forcing Model Intercomparison Project Instantaneous Radiative Forcing (RFMIP-IRF) outputs. Unlike cases where participating models run idealized cases and then seek to project those results back onto their model as it is run normally, we present native diagnostics whereby each participating model runs typical experiments and reports all fields necessary to calculate aerosol radiative forcing.

Comparisons of these results against a set of global benchmark line-by-line codes comprise a set of native model error diagnostics that provide novel insight into the sources of error, which can be decomposed into solver error, solar band error, and gas overlap error. Proof-of-principle from the GFDL AM4 and CESM1.2.2 models indicate that choices of radiative transfer solution can contribute to aerosol forcing error: the implementation of the two-stream approximation in the GFDL code is a primary driver of error for that model, while the representation of spectral bands is the primary driver for error in the CESM model.

For the latter case, diagnostics provide a wealth of geographic and spectral data, and they enable subsequent hypothesis-testing. These tests reveal that RRTMG’s errors can be ascribed to its coarse representation of spectral bands across visible and near-infrared wavelengths where aerosol optical properties change rapidly along with water vapor absorption.

While RRTMG has some mechanisms to correct for this by determining within-band angstrom exponent variation, those mechanisms may not be implemented uniformly across models. Given the widespread utilization of RRTMG, we run CESM with and without these mechanisms to quantify the implications for using the uncorrected code on the temporal and spatial evolution of aerosol forcing in the DECK experiments.

Link for More Information: