Testing an atmospheric model's radiative flux sensitivities at the top of the atmosphere
For the first time, scientists at Pacific Northwest National Laboratory conducted a comprehensive sensitivity study of 16 selected parameters in a popular atmospheric model to analyze their effect on the flux of energy at the top of the atmosphere. They found that cloud parameters—especially the one changing cloud ice to snow—are the primary culprits affecting energy flux among these parameters. They also found that pollution and natural emissions particles affect the atmosphere more at regional than global scales. Their analysis provides evidence that interactions among the selected parameters have little influence on the total mean net radiant flux in most global regions.
Improving the fidelity of atmospheric models is a key means to understanding interactions in the climate. Simulating the complex chemical and physical interactions in the atmosphere, to describe past and future atmospheric states, comes down to proven numerical stand-ins for various atmospheric processes. Climate modelers are working to narrow the broad range of possible answers in these calculations, a.k.a. uncertainty. In this paper published in Atmospheric Chemistry and Physics, they focused on model representation at the top of atmosphere, where solar energy enters the Earth's system and reflected and thermal energy depart. The balance or flux of those interactions determines the Earth's energy budget and is the main driver of surface temperature change. Improving atmospheric models' ability to depict this change ultimately will make climate predictions more compelling.
Due to the complexity of the Earth-atmosphere system, quantifying how radiative energy changes the climate has proven difficult and is limited by model and climate process uncertainties. The PNNL-led team performed the first comprehensive sensitivity analysis on the Community Atmosphere Model version 5 (CAM5) cloud microphysics and aerosol parameters.
The research team analyzed the sensitivity of net radiative fluxes at the top of the atmosphere to 16 selected uncertain parameters related to cloud microphysics and aerosols in CAM5. They adopted a modeling approach called quasi-Monte Carlo sampling to effectively explore high dimensional parameter spaces. They simulated output response variables using CAM5 for each parameter set, and then analyzed them using a generalized linear model. They also conducted a variance-based sensitivity analysis to show the relative contributions of individual parameters to perturb the global variance of the net radiative fluxes.
The results indicate that the changes in the global mean radiative flux are dominated by changes in net cloud forcing within the parameter ranges they investigated. The parameter for threshold size related to the auto-conversion of cloud ice to snow was identified as one of the most influential parameters for radiative flux in CAM5. Although anthropogenic and natural emissions largely affect radiative flux variance at the regional scale, their impact is weaker on the global mean radiative flux than the effect of the model's internal parameters. Finally, the interactions among the 16 selected parameters contributed a relatively small amount to the total radiative flux variance over most regions of the globe. This study provides better understanding of the parameter ranges in the CAM5 model, supporting further calibration of uncertain model parameters with the largest sensitivity.
This study indicates that reducing modeling uncertainty through calibration requires a complete understanding of the model behavior within the parameter uncertainties. Next, the team will analyze the precipitation and aerosol-cloud interaction modeling sensitivity of the CAM5 parameters.