Estimates of climate system properties based on recent climate records up to 2012

Thursday, 18 December 2014
Alex G Libardoni, Pennsylvania State University Main Campus, Meteorology, University Park, PA, United States, Chris E Forest, Earth and Environmental Systems Institute, The Pennsylvania State University, University Park, PA, United States and Andrei P Sokolov, MIT, Cambridge, MA, United States
We will present estimates of joint probability distributions for climate system properties that are based on new simulations using the MIT Earth System Model (MESM). Differing from earlier estimates, these simulations are based on updating the historical forcing data sets to include data through 2012. These input data include anthropogenic greenhouse gases, sulfate aerosols, land-use change, tropospheric and stratospheric ozone, stratospheric aerosols due to volcanic eruptions, and total solar irradiance. Additionally, the MESM has been updated to include a new land-surface model (the Community Land Model, CLM3.5) that provides an improved estimate of the components of the surface heat balance compared to CLM2.1. CLM 3.5 allows for an accounting of the impact of land-use change on surface characteristics, such as albedo and roughness.

We estimate the likelihood of the climate system properties using the approach of Libardoni and Forest (2013). We include model constraints based on changes in decadal mean, zonal-mean surface temperatures, upper-air temperature trends, and ocean heat content trends and include the period of record up to 2012. These results will be compared with Libardoni and Forest (2013) and results presented in the IPCC AR5 WG1. We will also investigate the dependence of parameter distributions to the structure of model diagnostics. Changes in the diagnostics include changing the end date of the period of record, changing the averaging periods, calculating means with respect to different climatology periods, spatial averaging schemes used to aggregate observational data to the model grid, and the spatial resolution of model diagnostics. In particular, reducing the spatial resolution of the upper-air diagnostic allows for better estimation of the natural variability and allows for temporal information to be accounted for when comparing model output to observations.

Further improvements to the likelihood estimates include updating noise-covariance estimates to use control run data obtained from CESM1/CCSM4 to be consistent across the multiple diagnostics. Using control run data from a consistent source allows for correlations and covariances in the model diagnostics to be accounted for in parameter estimation.