UC Berkeley

Global-scale atmospheric models play an important role in predicting atmospheric aerosol and the corresponding radiative forcing. Although atmospheric models are important tools, there is large uncertainty associated with aerosol predictions due to uncertainty in aerosol representation within the models. As a result, aerosols and their influence on the Earth's energy balance are considered one of the largest uncertainties in understanding climate change. Given the importance of simulating aerosol for understanding global climate change, it is evident that alternative methods are needed to reduce the effect of the uncertainties associated with aerosol representation and enhance the fidelity of the aerosol models. The goal of this research is to produce aerosol fields with reduced uncertainty by constraining model predictions with observations, using a technique known as data assimilation. The results from the aerosol assimilation are used to investigate aerosol sources, lifetime, and shortwave radiative forcing.

Two new aerosol data assimilations are presented as part of this work with both assimilations making use of an Ensemble Kalman Filter (EnKF) and the National Center for Atmospheric Research's (NCAR) community atmosphere model (CAM) with 60 ensemble members. The first assimilation involves the joint adjustment of the amount of atmospheric aerosol and the relative amount of fine and coarse aerosol using observations of aerosol optical depth (AOD) and angstrom exponent from NASA's Moderate Resolution Imaging Spectroradiometer (MODIS). Both the amount and relative contribution of fine and coarse aerosol were identified as key parameters for determining aerosol radiative forcing and, therefore, accurately determining these parameters is desirable. The second assimilation presented as part of this work is similar to the first with the addition of a vertical redistribution of coarse aerosol using vertical extinction observations from NASA's Cloud-Aerosol LIDAR and Infrared Pathfinder Satellite Observations (CALIPSO) satellite. Studies have shown that the atmospheric lifetime of aerosol is tightly coupled to the vertical profile, therefore, it is expected that vertical adjustments will further reduce aerosol uncertainty, especially in coarse aerosol. Similar to aerosol amount and size, lifetime is important for properly quantifying radiative forcing as it determines the time an aerosol has to impact the climate and influences the horizontal distribution of aerosol that is highly heterogenous in space and time.

The two presented assimilations are run for the year 2007 and results are compared against a control run simulation as well observations of AOD, angstrom exponent, and fine aerosol contributions from MODIS and NASA's Aerosol Robotic Network (AERONET). Through the comparison, it is demonstrated that the presented assimilations are able to reduce the model bias with an increase in the predicted aerosol optical depth. The globally averaged control run AOD prediction for 2007 is 0.086(± 0.06). The globally average AOD predictions for the amount and size assimilation and vertical assimilation are 0.115(± 0.05), 0.140(± 0.05), respectively. This is compared to globally averaged MODIS observations of 0.161(± 0.09). Over-ocean averaged angstrom exponent predictions from the control run are 0.65(± 0.35) while the size and amount and vertical assimilation predictions are 0.68 (± 0.15) and 0.66 (± 0.15), respectively. This is compared to globally averaged MODIS observations of 0.65(± 0.30). While it is difficult to determine improvements in angstrom exponent predictions based on the global average, clear reductions in regional biases were observed. Aerosol predictions are also compared to ground-based AERONET observations by site category, including desert dust, biomass burning, rural, industrial pollution, polluted marine, and dirty pollution. While the rural sites have statistically similar averaged AOD values across simulations, improvements are found for the other site categories in the assimilation runs with higher average AOD values and greater temporal variability. In addition to AOD comparison, the predicted amount of AOD due to fine aerosol is compared to AERONET observations by site category. The greatest reduction in bias is observed for polluted marine sites with the assimilation runs predicting a smaller fine aerosol contribution than the control simulation. Size-related observations are concentrated over ocean regions, therefore, the greatest impact of the assimilation with respect to size is expected for marine sites. Additionally, the positive bias in fine aerosol contribution predicted at dusty sites is reduced the most in the vertical assimilation with dust being mostly coarse in size. The adjustments to the vertical profile of coarse aerosol in the vertical assimilation further reduce bias for coarse dominated sites.

The results of the assimilation are used to quantify the contribution of anthropogenic aerosol to AOD. Globally averaged, the anthropogenic contribution to AOD is 38.8 percent for the control simulation, 47.6 percent for the amount and size assimilation and 49.5 percent from the vertical assimilation. These results are comparable to previously published anthropogenic AOD percentages which range from 41 to 72 percent (IPCC 2007). Additionally, aerosol loss processes and lifetime are analyzed. The dominant loss processes are condensational growth for nucleation mode aerosol (fine, < 0.1&mum), activation and wet deposition for accumulation mode aerosol (fine, 0.1-1&mum), and dry deposition and activation for coarse mode aerosol (> 1 &mum). The longest aerosol lifetimes are found in the vertical assimilation with most aerosol species showing better comparison to reported AEROCOM lifetimes, except for sulfate. In particular, the lifetimes of coarse mode dust and sea salt increased in the vertical adjustment assimilation, reducing the negative aerosol optical depth bias, especially in dusty regions. The predicted sulfate lifetime is double the reported AEROCOM value and may be the cause of some positive AOD bias regions in the Northern hemisphere predicted in the model runs.

The solar direct radiative forcing (DRE) is calculated using the predicted aerosol fields with the DRE including the effects of both anthropogenic and natural aerosol. Uncertainties in DRE for the assimilation runs are determined using the 60 member ensemble spread. Globally averaged DRE values are -1.9 W/m2, -5.2(± 0.51) W/m2, and -7.2(± 0.94) W/m2, for the control, amount and size and vertical assimilation, respectively. The predicted DRE from the amount and size assimilation compares the best to previously published estimates. Additionally, the calculated anthropogenic contribution to AOD is used in conjunction with the DRE estimates to calculate shortwave anthropogenic direct radiative forcing estimates with predicted values of -0.77, -2.3(± 0.64) and -3.2(± 0.7) W/m2 for the control, amount and size assimilation and vertical assimilation, respectively.