2014 Progress Report: Linking Regional Aerosol Emission Changes with Multiple Impact Measures through Direct and Cloud-Related Forcing Estimates
EPA Grant Number:
Linking Regional Aerosol Emission Changes with Multiple Impact Measures through Direct and Cloud-Related Forcing Estimates
Bond, Tami C.
, Amar, Praveen
, Liang, Xin-Zhong
, Streets, David G.
University of Illinois at Urbana-Champaign
Argonne National Laboratory
Clean Air Task Force
University of Maryland - College Park
University of Illinois at Urbana-Champaign
Argonne National Laboratory
EPA Project Officer:
August 1, 2011 through
July 31, 2014
(Extended to July 31, 2016)
Project Period Covered by this Report:
August 1, 2013 through July 31,2014
Black Carbon's Role In Global To Local Scale Climate And Air Quality (2010)
Global Climate Change
This project examines the effects of black carbon (BC) emissions on atmospheric concentrations, radiative forcing, and climate. BC absorbs solar radiation and therefore contributes to global warming. Because it is in particle form, BC also affects the formation and destruction of clouds, causing several secondary influences on climate. BC also can alter the albedo of the snow and ice surfaces. BC’s range of effects on the atmosphere and health suggests a rationale for control. The goal of this project is to move toward a framework for quantifying the effect of individual sources and source categories (such as diesel engines) on atmospheric impacts usually associated with climate. Because mitigation strategies can alter all emissions from a single source, especially particulate emissions, we also investigate the climate response to particulate organic matter (POM), which commonly is co-emitted.
Task 1: Expand emission model to include size information for primary aerosols and aerosol precursors
The objective of this work is to develop a global emission inventory with detailed PM size distribution, especially for submicron particles, and thereby to advance our understanding of the effects of particle size on global climate. The work began with an existing technology-based emission inventory containing about 400 separate technologies with regionally specific technology divisions. Particle emissions from technology then were combined with a literature review of particle size distributions. This work is now complete (Winijkul, et al., submitted). Major findings are: (1) for the global average, size-resolved PM emissions have a single mass distribution with the majority of particles in size ranges smaller than 1 µm (Figure 1); and (2) size distributions differ considerably by sector, with power-sector emissions being the largest in diameter, followed by industrial, residential, and transportation sectors in that order.
We also investigated three scenarios in which some of the most promising control measures were simulated in each of the industrial, residential, and transportation sectors (Figure 2). In the residential sector, when solid fuel is replaced with gaseous fuels, the size distribution shifts to a smaller size range with peaks at 20 nm and 200 nm. In the industrial sector, we replace cement emissions with baghouses, reducing PM10 emissions by 15% but PM1 by only 1%. In the transportation sector, all on-road vehicles are forced to follow Euro 6 emission standards, reducing emissions of PM smaller than 2.5 µm.
Task 2: Quantify direct radiative forcing and cloud radiative forcing by aerosols using CWRF and CAM-Chem with an optimized physics ensemble
CWRF-Chem (Regional Model)
In the original design of this project, a regional model was to be used with a Cloud-Aerosol Radiation (CAR) ensemble system. CAR has numerous choices of alternative parameterizations for cloud properties, aerosol properties, and radiation transfer (Liang, et al., 2013; Zhang, et al., 2013; Liang, et al., 2012). The purpose of using an ensemble system was to ensure that modeled radiation, which depends on the parameterization, matched the observed radiative fluxes. Aerosol forcing is the relatively small difference between an unperturbed radiative flux and a flux perturbed by aerosols, including the effect of aerosols on clouds. Thus, confirming that the clouds and the unperturbed flux are represented correctly is imperative before assessing a perturbation. However, the aerosols themselves also affect the modeled flux and the observed flux includes the effects of both clouds and aerosols.
We selected one cloud parameterization that gave a reasonable match for fluxes; during this year one parameterization within CAR was embedded in the mesoscale Climate–Weather Research and Forecasting (CWRF) model. The CAR ensemble system also allows selection of different aerosol properties, such as type, profile, and optics. In Year 2, we described modeling BC aerosols with CWRF directly, where BC concentrations were underestimated at the surface. Rather than devoting more time to tuning a model, we again opted for an ensemble approach. During Year 3, we conducted six experiments with aerosol distribution information imported from observations or models. Figure 3 shows aerosol effects averaged over the Continental United States (CONUS), where aerosol distributions were provided by MISR satellite, NCAR CAM2 model, NASA GOCART model, and NCAR CAM5 model show substantially different effects in CWRF radiation simulations. More details were included in Appendix B of our report submitted to EPA.
Clear-sky radiative changes are associated with direct forcing—the direct interaction of aerosols with solar radiation. Cloudy-sky radiative changes are associated with indirect forcing, or the change in fluxes caused by the alteration of cloud properties. However, some cloudy-sky radiative changes may be part of direct forcing if they occur over clouds. Figure 3 shows that in many models, the net flux is the sum of a positive (clear-sky) and a negative (cloudy-sky) flux. These large counterbalancing effects are different than observations. Work in the coming year will focus on diagnostics to isolate the physical causes of the differences among models, especially with the older CAM2 model that better matches observations. In turn, this will lead to an understanding of why models predict different values for aerosol forcing.
Many of the model experiments examine the effects of total aerosol loading, instead of BC alone, because of the need to compare with observations. We also conducted case studies to investigate the attribution of forcing to BC aerosols. With the same total aerosol loading, we altered 15% of BC from hydrophilic to hydrophobic, and found substantial changes (± 5 W/m2) in the cloud radiative effects (Figure 4). These effects are regionally dependent, with positive effects of up to 10 W/m2 in the southeastern United States, and negative effects of 4-6 W/m2 in the northeast coastal areas. Again, work in the remaining year will seek diagnostics to understand how aerosol distribution influences the simulation results.
CAM 5.2 (Global Model)
Modeled forcing was reported in Year 2 for the global Community Atmosphere Model (CAM). Activities with CAM this year focused on exploiting simulations to determine policy-relevant metrics and are reported in the next section.
Task 3: Test emission-to-forcing relationships that capture source-receptor regional impact
Radiative forcing is commonly presented as the Earth’s response to a single species. Assuming that one can reduce this forcing by eliminating emissions from a source or group of sources implies that forcing or climate impact is proportional to emission. Emission metrics such as the Global Warming Potential also implicitly assume this proportionality between forcing and climate impact. This linearity is expected to be valid for direct forcing, but the proportionality has not been tested. Cloud-related forcing is known to be nonlinear with regard to emission, with the effect becoming saturated at high aerosol loadings. The validity of assuming relationships between forcing and emission have not been tested.
Using the global CAM model, we studied the relationship between forcing and emission in simulations where BC or POM emissions were reduced from the two regions identified in the proposal: North America and Asia. We depart from typical forcing treatments by examining radiative changes from present-day rather than pre-industrial conditions, because the present-day will serve as the baseline for any mitigation actions. The results of the CAM simulations are summarized here.
Using original emission as the base case, a measure of linearity is the ratio (R) of forcing when emission is reduced to 50% and the forcing when emission is reduced to 0. A value of 0.5 indicates that forcing is perfectly linear in emission. Values much lower than 0.5 (“sublinear”) indicate that the first increment of emission reduction has a relatively lower effect on forcing, as might be expected if the indirect effect is saturated. Figure 5 summarizes these values for BC and POM in two regions, separately for the direct and indirect effect. The indirect forcing of POM shows sublinearity with forcing, with R equal to 0.39 and 0.31.
Forcing mainly occurs close to the source region, yet most model presentations give global averages. We experimented with regional sizes for aggregating impact, using principles of tested for statistical significance, because forcing is a small signal (change) in a noisy system. We found that the optimal aggregation size is 30 degrees by 30 degrees, so the globe is covered with 72 boxes. We found that significant regions are 50% or less of the total boxes, but these boxes contain more than 75% of the global average forcing. This means that global average forcing can be constrained by examining a much smaller number of near-source grid boxes. With these significant regions in mind, we investigated linearity in forcing-to-emission relationships at regional scales. While direct forcing is always linearly related to emission, substantial sublinearity was found for indirect forcing, especially in regions with high forcing values that are near the source region.
The cause of sublinearity between emission and indirect forcing was studied by decomposing the physical process from mass emission to indirect forcing, which includes the relationship between aerosol number, cloud condensation nuclei (CCN), cloud droplet number and forcing. The modeled sublinearity occurred at the intermediate step from CCN to cloud droplet. Changes in emission linearly affect aerosol number and CCN, but not all CCN activate into cloud droplets, probably due to depletion of liquid water. While this result could be expected because of the model parameterization, it has not been confirmed that this is the dominant effect and the extent to which nonlinearity occurs has not been quantified.
Task 4: Distill and communicate quantification measures that capture direct and cloud-related forcing
During this year, we finalized the report on a first round of surveys, which now has been communicated to EPA. Our original proposal had envisioned working with state-level policymakers to evaluate communicating the science of black carbon and climate and, more generally, aerosols and climate. Because of political changes during the course of the proposal, the topic of climate became less welcome in state-level discussions. However, climate considerations did become a component of action plans for large cities. We designed a new survey to isolate the factors that are important to city-level planners. The rate of response has been poor and our final year’s work will require exploring incentives for participation in this short survey.
The main challenge in this project is the level of integration compared with what was originally conceived in the proposal. The size-resolved global emission inventory is complete, but we found that models are not yet ready to implement a size-varying inventory, despite claims that they include aerosol size. Comparisons between the regional model and the global model are finally underway, yet the two models have fundamentally different assumptions and ways of storing and presenting data. Community-wide standards for model storage are needed.
A major result from this project is increased confidence in the use of a linear emission-to-forcing relationship for cloud-related forcing, especially for North America. This confidence is required to develop simple policy-relevant measures that reflect total climate impact.
Emission model: None (complete).
Radiative forcing, CWRF: Diagnose causes for differences among models and with observations; complete publications.
Radiative forcing, CAM: Prepare publications describing linearity of forcing; continue to explore diagnostics for causes of differences between regions; use model experiments to investigate validity of combining forcing for component emissions (such as BC and POM) to estimate forcing from a single source. Explore indirect forcing by elevated versus surface emissions.
Communication group: Explore incentives to obtain responses from city environmental managers.
Liang XZ, Zhang F. Cloud-Aerosol-Radiation (CAR) ensemble modeling system. Atmos. Chem. Phys. Discuss. 2013;13:10193-10261.
Liang XZ, Xu M, Yuan X, Ling TJ, Choi H, Zhang F, Chen LG, Liu SY, Su SJ, Qiao FX, He YX, Wang XL, Kunkel K, Gao W, Joseph E, Morris V, Yu T, Dudhia J, Michalakes J. Regional Climate-Weather Research and Forecasting model. Bull. Amer. Meteorol. Soc. 2012;93:1363-1387.
Zhang F, Liang X-Z, Li J, Zeng Q. Dominant roles of subgrid-scale cloud structures in model diversity of cloud radiative effects. Journal of Geophysical Research: Atmospheres 2013;118:7733-7749.
Winijkul E, Fan F, Lu Z, Streets, DG, Bond TC, Zhao Y. Size-resolved global emission inventory of primary particulate matter from energy-related combustion sources. Atmospheric Environment 2015;107:137-147. (noted as submitted in report text)
on this Report
: 1 Displayed | Download in RIS Format
Progress and Final Reports:
2012 Progress Report
2013 Progress Report
2015 Progress Report