California Air Resources Board

Cancer Inhalation Risk Modeling Method Summary

Page Reviewed July 28, 2010
Return to -About Risk Maps- Page

This page briefly explains the methods used for creating the cancer inhalation risk maps. Core modeling issues appear toward the top of the list. Near the bottom of the listing are discussions relating the ASPEN model to real-world data, and to other toxics modeling approaches.

Toxic dispersion model used: ASPEN 

Dispersion modeling was conducted with ASPEN, the "Assessment System for Population Exposure Nationwide." The ASPEN Model source code used by the ARB was identical to that used by the U.S.EPA for their 1990 through 1999 inventory year National Air Toxics Assessment (NATA) ASPEN model runs. ASPEN is a peer-reviewed modified version of U.S.EPA's Human Exposure Model (HEM), which models long-term concentrations for simple terrain over large areas. The ASPEN model incorporates improvements to the HEM, including improved reactive decay, secondary compound formation, and particulate matter deposition.

Risk calculated using ASPEN 

Cancer inhalation  risk factors have been established by peer review, involving a broad range of stakeholders. These factors are applied to the ASPEN modeled ambient concentrations to calculate the toxic risk. With the exception of the unit risk factor used for the calculation of the compression ignition diesel risk, all risk is calculated at the individual compound (e.g., benzene) or grouped compound (e.g., hexavalent chromium) level. These risks are then summed to determine composite risks.

Emissions inventory used 

Nearly all of the inventory reflected in the maps was the 2001 CEIDARS inventory, frozen in September of 2003. There have been changes in the inventory between the previously posted 2000 baseline inventory year (frozen September of 2002) risk maps, and these currently posted 2001 baseline inventory year risk maps. However, most of the changes have not resulted in significant adjustments to the results of the ASPEN-based risk mapping. The change with the largest impact was a small increase in the offroad diesel inventory, due to additional equipment categories being included in the inventory.

The individual compounds shown in the inventory were derived either by speciating the TOG (total organic gas) and PM10 (particulate matter less than 10 microns) data in the CEIDARS criteria pollutant inventory, or, alternatively, from the AB2588 toxic inventory data stored in the CEIDARS data base. The only non-CEIDARS emissions data used was for the Los Angeles region marine shipping categories. That data was derived from 1997 shipping data presented as moving point sources, which was then scaled to year 2001 levels. This was done because of the more precise, and detailed nature of the 1997 study data.

Meteorological inputs 

The ARB ASPEN model run used meteorological files produced by the U.S.EPA for their 1999 ASPEN model run. The 1999 meteorology will continue to be used by ARB staff for ASPEN modeling, regardless of modeled year, until an improved meteorological approach is determined. ARB staff determined that generating a new year 2001 meteorological data set for this risk map data set would not yield significant improvements to the lifetime inhalation cancer risk estimates. Unlike ozone episode modeling, where the purpose is to determine the impacts of emissions sources on peak days, the inhalation cancer risk maps shown on this web site are for lifetime exposure at the emissions levels for the year modeled. The ideal scenario would be to model every year for the 70 year lifetime. However, there is no way to determine future weather with any accuracy. Therefore, a representative year is being used. The limitation to this approach is that comparisons of ASPEN modeled concentrations to ambient monitor data will be more difficult, since the year 2001 meteorology is not being matched to the year 2001 inventory. The benefit of this approach is that it makes comparisons between different modeled years easier, because meteorological influences can be eliminated.

Time scale of ASPEN model 

ASPEN models the entire year as a single day, split into separate three-hour timeblocks. This approach averages the emissions across the entire year for each 3-hour timeblock. After the ASPEN model run has been completed, the eight 3-hour timeblocks are averaged to create the annual census tract level concentration.

Geographic resolution of ASPEN model 

ASPEN calculates concentrations of air toxics for each census tract. Typical census tracts may contain about 4,000 residents, although this number can vary significantly. Urban census tracts are often about 2 square miles in size, while they will be much larger in rural areas. The Aspen model is designed for relatively large areas, and is of limited use for targeted small areas, like single facilities or local neighborhoods.

Region of emissions source impact 

The ASPEN model estimates outdoor air concentrations for a 50 kilometer radius around the sources. 

Spatial distribution of nonindustrial, mobile, and some industrial sources  

For all nonindustrial (solvent/other), mobile sources, and some industrial sources, surrogates (e.g., population, employment,vehicle activity, etc.) are used to assign emissions to modeling domain grid cells. The grid cell emissions are then apportioned to census tracts. 

Reactive decay 

The ARB ASPEN run incorporated the same reactive decay tables as the U.S.EPA Aspen Model.

Secondary pollutants 

The ARB ASPEN model run incorporated the same basic chemistry, and factors which the USEPA used in calculating their secondary pollutant concentrations for compounds such as formaldehyde and acetaldehyde. 

Particulate matter deposition 

The ARB ASPEN model run incorporated both fine and coarse particle deposition.

Background concentrations (not used in risk maps) 

For a number of pollutants, there are sources that are not a part of the emissions inventory that must be incorporated into the model as background pollutant concentrations. Ambient pollutant concentrations can result from sources outside of the modeled domain and/or persistent ambient concentrations from time periods prior to the modeled year; or from natural emissions that are not a part of the modeled inventory. No background pollutant levels have been included in the calculations of the risk values shown on the ARB's gridded risk maps. This was done so that the maps would be more useful in determining the local sources contributing to risk. 

ASPEN and human exposure  

There are ongoing efforts to better describe the actual human exposure to ambient toxic pollutant concentrations. Some studies use surveys, and personal logs kept by individuals to attempt to create a statistical sample of human activity in modeled regions. As the data, and appropriate human behavior modeling approaches become available, the ASPEN-based risk mapping approach could be adjusted.

ASPEN results versus monitor data 

Both the U.S.EPA and the ARB ASPEN model results have been compared to monitor data, and the modeled concentrations are generally within a reasonable range of actual concentrations (2 to 3 times). For both the U.S.EPA and the ARB ASPEN model runs, the modeled concentrations tend to be lower than the monitor data. These conclusions are based on a limited number of comparisons, and additional compounds need to be compared to a larger database of monitor data.

Underestimation of monitored data  

The USEPA has suggested that the tendency for underestimation is due to limitations in the ASPEN model related to such things as stagnation and nighttime stability. They also indicate that the 50 kilometer cutoff inherent in the ASPEN model, which limits source to target impact, may also be a contributing factor. 

Top of Page
Please send questions or comments to:

A department of the California Environmental Protection Agency