I. Overview

The Diesel Emissions Quantifier Benefits Module is a tool for estimating the health and monetary benefits that could result from a decrease in diesel exhaust emissions. The Benefits Module is a new component of EPA’s existing web-based Diesel Emissions Quantifier (the Quantifier). The Benefits Module uses the 2002 National Emissions Inventory (NEI) data and the 2002 National Air Toxics Assessment (NATA) model results to estimate the relationship of changes in diesel emissions to changes in primary particulate matter air concentrations for each county in the U.S. The Benefits Module then uses previously-generated outputs from the Environmental Benefits Mapping and Analysis Program (BenMAP) model to estimate the value of changes in the incidence of avoided premature mortality and several excess morbidity endpoints.

II. Estimating Changes in PM_2.5 Air Quality Concentrations Resulting from Diesel Emissions

A. Data Inputs

i. National Emissions Inventory

The NEI is a comprehensive inventory covering criteria pollutants and hazardous air pollutants (HAPs) for the 50 states, Washington DC, Puerto Rico, and the U.S. Virgin Islands. The NEI is assembled and reported every three years by EPA’s Emission Inventory and Analysis Group.  

Sources in the NEI are described as either stationary or mobile sources. Mobile sources are categorized as either on-road or non-road sources. On-road sources include motorized vehicles that are normally operated on public roadways. This includes passenger cars, motorcycles, minivans, sport-utility vehicles, light-duty trucks, heavy-duty trucks, and buses. Non-road sources include recreational marine and land-based vehicles, farm and construction machinery, industrial, commercial, logging, and lawn and garden equipment, aircraft, airport ground support equipment (GSE), locomotives, and rail maintenance equipment. These sources are powered by diesel, gasoline, compressed natural gas (CNG), and liquefied petroleum gas (LPG) -fueled engines, among others. 

ii. National-Scale Air Toxics Assessment

B. Analysis and Calculations

The census tract level NATA-predicted ambient concentrations of diesel PM were used to create the lookup tables that are the basis for the Benefits Module. These predicted ambient concentrations of diesel PM are used in conjunction with standard PM_2.5 concentration-response functions used in the BenMAP benefits modeling tool.

In order to create county-level ambient concentrations, the census tract level ambient concentration NATA results (c) have been population-weighted to county values:

For a county i and tract a:

This analysis was performed for total ambient diesel PM, as well as for on-road diesel PM and non-road diesel PM. County-level, versus census-tract level, concentrations have been used for the Benefits Module because (a) county-level results are a better match for the standard PM_2.5 concentration-response functions used in the health benefits analysis and (b) mobile source emissions, taken from NEI, are estimated at the county-level and the use of census tract level results would introduce additional uncertainty.

As an additional note, long range dispersion of diesel PM may contribute to an increase in diesel PM concentrations in one county due to emissions from a neighboring county. In general, this effect is likely to be insignificant because the large majority of diesel impacts occur in close proximity to the source, but it is a potential concern for low-emitting counties in close proximity to or downwind of one or more high-emitting counties. In areas where long-range transport is more important, the uncertainty resulting from the approach used here may be more significant and remains inadequately accounted for in this methodology. An inspection of the resulting county ratios, described in the Quality Assurance section below, reveals that there were only a few counties that had very low emissions and large ambient PM diesel concentrations, none of which were clearly an inaccurate result. Nonetheless, given the uncertainty in the results for these counties, benefits have been calculated, but a flag has been added in the Benefits Module to indicate where benefit per ton estimates for low-emitting counties may be underestimates, and also where benefit per ton estimates for high-emitting counties may be overestimates (due to transport of emissions into surrounding counties).  The method used to identify and flag counties, based on a ratio of predicted ambient concentration to emissions density in each county, is described further in Section IV.C below.

III. Estimating the Human Health Benefits of CHanges in PM_2.5 Air Quality

 A. Overview

Having first estimated change in PM_2.5 ambient concentrations resulting from a change in diesel PM_2.5 emissions, the Benefits Module then estimates the per-ton benefit of reducing ambient diesel PM_2.5. To perform this benefits analysis, the Benefits Module uses the “damage function” approach, which is a peer-reviewed technique for estimating the human health impacts associated with exposure to ambient pollutants (Levy et al., 1999). As a result, the Benefits Module calculates the benefit-per-ton of PM_2.5 emission reduction in a manner generally consistent with the methods found in the recently published Regulatory Impact Analysis (RIA) for the Ozone NAAQS (U.S. EPA, 2008a).

Estimating PM_2.5 benefit-per-ton entails three basic steps:

·      Estimating the change in PM_2.5 air quality for the geographic area of interest

·      Loading the estimated air quality changes into the Environmental Benefits Mapping and Analysis Program (BenMAP) and estimating the resulting change in the incidence of health outcomes and monetizing the benefits of those outcomes (Abt Associates Inc., 2005a)

·      Dividing the total monetized benefit by the total estimated emission reduction

B. Data Inputs and Health Endpoints

The Benefits Module uses the BenMAP model to estimate the health endpoints (the health effects that are caused, exacerbated, or otherwise affected by exposure to PM_2.5 such as premature mortality or asthma attacks) resulting from a unit change in diesel emissions in each county. Table 1 below summarizes the health endpoints quantified and the health impact functions applied for this analysis.

Modeling was done for each of three air quality modeling scenarios — on-road, non-road and total diesel PM. The model compared baseline air quality for each scenario (reflecting total county level ambient PM_2.5 from that particular source type alone) and a control air quality scenario (reflecting a zero-out of ambient PM_2.5) The modeling predicted relatively small incremental changes in PM_2.5 in each county. Because most of the health impact functions (equations that explain the relationship between exposure and changes in health endpoints) used for our analysis are log-linear (and thus produce different estimates of health impacts depending on the baseline level of air quality change), the benefits are somewhat sensitive to the baseline levels of air quality. For this reason, we modified the air quality inputs slightly by adding 10 µg/m³ to the baseline and control air quality files — ensuring that the benefits were calculated higher on the curve. Because it is not possible to know ex-ante what the baseline air quality levels will be in the counties in which users apply the benefit-per-ton estimates, this seemed like a reasonable adjustment.

In general, the benefits assessment used techniques, health impact functions and valuation functions that are consistent with the PM_2.5 health impacts assessments supporting the PM_2.5 and the Ozone National Ambient Air Quality Standards (U.S. EPA, 2006; U.S. EPA 2008a), with two major exceptions. First, in contrast to those analyses, this assessment applies non-threshold adjusted PM_2.5 health impact functions. Some researchers have hypothesized the presence of a threshold relationship between PM_2.5 exposure and the risk of adverse health effects, including premature mortality. For this reason, EPA has traditionally applied an assumed 10 µg/m³ cutpoint to the long-term mortality and short-term morbidity concentration-response functions. We determined that such a threshold would be inappropriate for this analysis because we do not know, ex ante, which areas would receive air quality improvements above or below this hypothesized threshold. Further, we did not believe it appropriate to assign zero benefits to counties where ambient PM levels were below a threshold level of 10 µg/m³.

The second major divergence from the two RIAs noted above is that we estimated current year population exposure (2008), rather than a projected exposure. We anticipated that most users would wish to estimate the near-term benefits of diesel control strategies, which called for using current year population to generate exposure estimates in BenMAP. Users interested in additional details regarding the health impact assessment may refer to the most recent PM_2.5 and ozone RIAs (U.S. EPA 2006; U.S. EPA 2008a).

Table 1: Summary of health endpoints and health impact functions

^a     The original study populations were 8 to 13 for the Ostro et al. (2001) study and 6 to 13 for the Vedal et al. (1998) study. Based on advice from the SAB-HES, we extended the applied population to 6 to 18, reflecting the common biological basis for the effect in children in the broader age group.

i. Annual versus Annualized Monetized Benefits

The steps above produce an annual estimate of the benefits of reducing an incremental ton of PM_2.5 from various emission sources for the year 2008. However, we expect that diesel retrofits  will provide a stream of benefits over a number of years. Moreover, the costs of these controls are frequently expressed in annualized terms that take into account the expected “lifetime” of the investment. Annualizing costs is the process of combining capital and operating-and-maintenance costs and then distributing these costs on an annual basis over the life of the equipment.

Thus, the benefits and costs are expressed in somewhat different temporal scales. Ideally, the benefits should also be annualized as well. However, this process would require a year-to-year estimate of the change in emissions and air quality over the life of each piece of equipment. For this same time period we would calculate year-to-year benefits, and this stream of future benefits would then be discounted back to the original year in which the emission control was installed. Moreover, we would account for year-to-year changes in population growth and distribution. We would also project changes in income growth to account for the increasing willingness to pay to reduce mortality risk. These were not practical analyses for this project.

Instead, we have made the assumption that the annual benefits are a fair surrogate for the annualized benefits. On one hand, we have neither modeled future population growth and distribution, nor accounted for future income growth; these are factors that should increase benefits over time. On the other hand, this stream of benefits would be discounted, which would reduce the annualized benefits. In our judgment, these countervailing factors more or less balance out such that the annual benefits are comparable to the annualized benefits. Each of the tables and maps in this document treat annual benefits as annualized benefits.

IV. ESTIMATING ANNUALIZED COSTS

The Quantifier estimates the cost-effectiveness of each project over the average remaining lifetime of the engine. These values are not easily comparable to the annual benefits presented in the Benefits Module. Therefore, the Benefits Module also estimates the annualized cost of each project.

The annualized cost is based on project cost data the user inputs into the Quantifier. Users can enter two different costs into the Quantifier: the total project cost and the capital costs. The total project costs refer to the entire cost of a retrofit project (for example, the amount of grant funding received to do the project) whereas the capital costs refer to the portion of the total costs that go towards purchasing and installing the retrofit equipment. Capital costs do not include any on-going maintenance costs. To calculate the annualized cost, the Benefits Module uses the value the user enters for the capital cost of the project.

The formula for calculating annualized costs in essence “spreads out” the initial investment costs of the project over the remaining lifetime of the engine being retrofitted. The remaining lifetime is calculated from the existing scrappage tables in the Quantifier. These are the same data used to calculate the cost-effectiveness estimates in the existing version of the Quantifier. This process is used because although the costs are usually paid upfront, the benefits are spread out each year over the remaining lifetime of the engine. By annualizing costs and benefits, the values can be more easily compared.

The formula used for annualizing costs is:

AC = (P * r)/(1-(1+r)^-n)

Where:

AC = Annualized Cost

P = Principal (or upfront capital cost)

r = Discount rate

n = Years (remaining life of the engine)

V. Uncertainties, limitations, and Quality Assurance

The Benefits Module represents a new way to bring together existing tools and databases to provide information to state and local agencies, the public and other parties as they seek to implement diesel reduction strategies. These existing data and tools have at various times been subjected to comment and peer review and reflect the recommendations of many experts in multiple disciplines. Nonetheless, the approach and data used by the Benefits Module contain multiple uncertainties and limitations that can limit the application of this tool. These uncertainties and limitations are discussed in more detail below.

B. Appropriate Use of This Application

For all of these factors, the uncertainty may lead to either a positive or negative bias in the results. The potential magnitude of the uncertainty in results is difficult to quantify. Past experience with emissions inventories would suggest that the magnitude of emissions, a product of emissions factors and activity, would be one of the largest uncertainties associated with the use of these data. However, basing our estimate on the ratio of the ambient concentration to total emissions, as is done for the Benefits Module, tends to minimize the importance of uncertainties in the emissions. For example, doubling emissions in a specific area would tend to double ambient concentrations, but keep the ratio relatively static, and thus the absolute uncertainty in emissions is not as significant a concern as other uncertainties in this analysis. Conversely, to the extent that these emissions transport to other areas, the uncertainty may be larger.

One of the main factors determining magnitude of health benefits associated with a given emissions reduction is the proximity of the emissions to people. Thus, uncertainty in the apportionment of emissions could be an important factor in this analysis. There are two things to consider for this uncertainty. First, if emissions are assigned to a larger census tract, then the same level of emissions will result in a lower ambient concentration, on average (the pollution, in effect, being spread out over a larger area means that there is less of it at any given point in that area). The opposite is true as well (i.e., assigning emissions to a smaller census tract will result in higher average concentrations). Second, if emissions are assigned to a less populated census tract, fewer people will be exposed to the resulting concentration of air pollution and the population-weighting at the county scale will predict a lower concentration and thus a lower ratio. Again, the opposite is true (i.e., emissions assigned to higher-populated tracts leads to an overestimate of concentration and ratio).

We do not anticipate a high degree of uncertainty associated with treating mobile sources as a series of radial points within census tracts, although this may be more of a concern for counties and census tracts that cover a large geographic area. The Benefits Module uses average concentrations at a much larger geographic scale (i.e., county-level), which would tend to underestimate the importance of local hotspot impacts that are not detected by the NATA approach. Some bias may result, however, if the population within a census tract is located closer to and therefore more exposed to pollution from major roads or other low-level releases than our analysis assumes.

The health benefits in the Benefits Module are for PM_2.5 generically and are not dependent on the precise chemical composition of the PM_2.5 emissions in a particular area. Therefore the only likely significance associated with not considering atmospheric chemistry is if chemical reactions could lead to either loss or formation of PM_2.5. The loss of directly-emitted diesel PM through chemical reactions is unlikely, since the impacts from diesel PM tend to be highly local for these source types (e.g., no high stacks, minimal exit velocity) and there is insufficient time for reactions to occur before concentrations have been diluted by dispersion alone. Dilution of diesel PM occurs in less than 1 mile, or less than 20 minutes at even slow wind speeds, which is much faster than the typical atmospheric half-life of PM_2.5, which is considered to be on the order of days to weeks (e.g., Wilson and Suh, 1997).

In addition, the exposure and benefit-per-ton values do not include highly localized exposures, such as those that occur when diesel exhaust “self-pollutes” the cabin on the vehicle from which it has been emitted. This phenomenon has been studied extensively in diesel school buses, and the data indicate it can be a significant source of exposure from older diesel engines(e.g. Marshall and Behrentz, 2005). This Benefits Module does not capture this type of micro-scale exposure and thus the benefits estimate does not include the benefits of reducing these types of exposures.

Uncertainties in the use of NEI emissions and NATA-predicted ambient concentration may be reduced by considering the following when calculating health benefits using the Benefits Module:

·      The highest uncertainties in the Benefits Module’s emissions, dispersion, health, and monetary benefits calculations are likely all associated with considering only a single location or project. Uncertainties that may have either a positive or negative bias when considered together are more likely to be substantially smaller when considering multiple emissions reductions over larger geographic areas, to the extent that such bias is not highly correlated with population.

·      The results of the Benefits Module may be used to characterize the relative benefits of diesel emission reduction projects between areas, but comparisons are likely to be more uncertain when comparing areas in different states, where differences in underlying methodology (e.g., local submission of emissions information to NEI) are likely to be more significant.

·      The benefits module is most appropriate when used to estimate scale and relative distribution of results (as opposed to precise predictions) and thus should be used for purposes where this type of estimate is appropriate only. These results are not an adequate substitute for a more refined emissions, dispersion, and health impacts analysis in support of broader decision-making.

Both the calculation of air concentrations from emissions estimates and the subsequent estimation of the health benefits of those improvements in air quality are subject to significant uncertainty. As stated earlier, these estimates should be considered just that: estimates, and not precise calculations or predictions.

C. Quality Assurance

Figure 1 and Tables 2 through 4 are designed to examine whether the highest predicted benefit-per-ton results are reasonable. One of the primary concerns with our methodology is with counties that may experience substantial diesel impacts due to atmospheric transport from surrounding counties, but may not themselves have substantial emissions. This would likely skew the results towards unusually high benefit-per-ton numbers in those counties (i.e., skewed higher ratios of NATA-predicted diesel PM concentrations versus county emissions would be used as inputs for benefits calculations in BenMAP).

Figure 1 is a plot of monetary benefit-per-ton of diesel emissions reduced (expressed in $/ton) for each county in the United States versus total emissions (tons/year), by source, in that county. This figure illustrates two main points. First, there are few, if any, outliers with high benefit-per-ton but low local emissions. Although this figure cannot illustrate sufficiently whether the low-emitting counties are nonetheless skewed higher by atmospheric transport than would otherwise be expected, no low-emitting counties have benefit-per-ton results beyond what is observed for higher emitting, and thus more certain, counties. Second, the distributions show a relative positive trend; that is, benefit-per-ton estimates increase with county emissions. This result is reasonable because higher emitting counties also tend to be more populated counties and the combination of a higher density of sources and population in proximity to each other would lead to higher anticipated health benefits for diesel exhaust reductions.

Another way to consider the impacts of atmospheric transport either into or out of a county is to estimate the import/export factor. This factor describes the relationship between the change in NATA-predicted ambient concentration to the change in emissions density for that county.  Figure 2 shows a plot of monetary benefit-per-ton of diesel emissions reduced (expressed in $/ton) for each county in the United States versus the ratio of change in concentration versus change in emissions density. This can be indicated by. Dc_i/(De_i/a_i), where c_i, e_i, and a_i are the concentration, emissions, and area of county i.  Counties that are highest in Dc_i/(De_i/a_i) would be indicative of those that are most likely to import a relatively large portion of diesel PM, while counties that are lowest in Dc_i/(De_i/a_i) would be indicative of those that are most likely to export a relatively large portion of diesel PM. A high import/export value indicates the air concentrations in the county are likely affected by imports of diesel PM from other counties. A low value indicates the county is likely to export a large portion of the diesel PM emitted there to other counties.

Figure 1:  Monetary benefit-per-ton of diesel emissions reduced ($/ton) for all counties in the United States plotted versus source-specific diesel emissions (tons/year) in that county. Results are presented for (a) total diesel sources, (b) on-road diesel sources, and (c) non-road diesel sources.

Figure 2:  Monetary benefit-per-ton of diesel emissions reduced ($/ton) for all counties in the United State versus the import/export factor. This factor is a ratio of change in concentration versus change in emissions density for each county, i.e., Dc_i/(De_i/a_i), where c_i, e_i, and a_i are the concentration, emissions, and area of county i.  Results are presented for (a) total diesel sources, (b) on-road diesel sources, and (c) non-road diesel sources.

Tables 2 through 4 show the counties with highest predicted benefit-per-ton due to reductions from total diesel sources, on-road diesel sources, and non-road diesel sources, respectively. Tables 5 through 7 show the benefit-per-ton for counties with the lowest emissions for total diesel sources, on-road diesel sources, and non-road diesel sources, respectively. Tables 8 through 10 show the counties with the highest import/export factor (i.e., counties likely to import) and Tables 11 through 13 show the counties with the lowest import/export factor (i.e., counties likely to export) nationally.

A closer examination of the counties with the highest-predicted benefit-per-ton estimates (Tables 2 through 4) shows that counties with a high density of sources and/or high population density (such as Bronx, Kings, New York,  Manhattan, and Queens Counties, which are part of the City of New York) have some of the highest benefit-per-ton estimates, which is expected. The independent cities of Virginia, i.e., Fairfax, Poquoson, Portsmouth, Winchester, Franklin, Lexington, and Falls Church, also show very high benefit-per-ton results, especially relative to their local emissions. These results do not appear unreasonable since these cities tend to be fairly dense with both sources and receptors. Many of these same counties have the lowest import/export factors in Tables 11 through 13, supporting the assertion that, if anything, the counties are mostly exporters of diesel emissions and the benefit-per-ton estimates may be underestimates.  

Most of the instances of unusually high or low benefit-per-ton results are for non-road emissions. For example, the Loving County, TX, benefits of $42,000 per ton (Table 7), while small, is most likely due entirely to transport of outside pollutants, because there are essentially no local sources. Similarly, the $520,000 per ton for Alpine County is quite large, given the minimal local sources (0.3 tons/year) and sparsely populated, low density county. The import/export factor analysis supports both of these assertions, since both counties have a very high ratio (Table 13), and could thus be interpreted as diesel importers.  Two other counties with low emissions (<2 tons/year) but high predicted benefit-per-ton (> $500,000 per ton) are Owsley County, KY, and Clay County, WV.                                         

In order to acknowledge this uncertainty, the diesel benefits calculator takes the following approach. First, in addition to reporting results for the county selected, the results are also calculated and reported using statewide benefit-per-ton values in order to provide context. Second, for all counties with import/export factors in the lowest 5^th percentile – for either on-road or non-road sources, depending on the query – the results are flagged with the following message:

Benefits estimates are “flagged” for this county, indicating that we have less confidence in these results due to a large amount of inter-county transport of emissions. The impacts estimation tool may be underestimating benefits for emissions reduction projects in this county because it has a relatively high density of emissions compared to surrounding areas. As a result, this county is likely to be a net exporter of diesel emissions, and many of the benefits of reducing these emissions are likely to take place in downwind counties. Please take this increased uncertainty into account when interpreting your results.

Also, for all counties with import/export factors in the highest 5^th percentile – for either on-road or non-road sources, depending on the query – the results are flagged with the following message:

Benefits estimates are “flagged” for this county, indicating that we have less confidence in these results due to a large amount of inter-county transport of emissions. The impacts estimation tool may be overestimating the benefits for emissions reduction projects in this county because it has relatively few emissions compared to surrounding areas. As a result, this county is likely to be a net importer of diesel emissions, and air quality is significantly affected by emissions in upwind counties. Please take this increased uncertainty into account when interpreting your results.

EPA also calculated a population-weighted average of the county benefit-per-ton values within each state and within the entire United States. The procedure was identical to the population-weighting performed for averaging census tract ambient concentrations to the county level.

For total diesel sources, we calculated a range from $3.2 million per ton for New York State to $68,000 per ton for Wyoming. The national population-weighted average was $1.2 million per ton. The national benefit-per-ton value is somewhat higher than the national mobile source benefit-per-ton from carbonaceous particles from all mobile sources of $730,000 that was calculated as part of the ozone NAAQS RIA (U. S. EPA, 2008a). For on-road diesel sources we calculated a range from $3.8 million per ton for New York State to $63,000 per ton for Wyoming. For non-road diesel sources we calculated a range from $3.2 million per ton for New York State to $73,000 per ton for Wyoming. The national population-weighted average for on-road sources and non-road sources are $1.2 million per ton of diesel reduced. This is also somewhat higher than the on-road and non-road estimates calculated as part of the ozone NAAQS RIA, which are $740,000 per ton and $720,000, respectively.

The benefit-per-ton estimates from this project are clearly very different from those in the most recent ozone RIA. However, the divergence may be due to the fact that the diesel PM benefit-per-ton estimates reflect air quality changes from diesel sources alone. Conversely, the benefit-per-ton estimates developed for the ozone RIA reflect air quality changes from reductions in carbonaceous particles across all on-road and non-road mobile sources. Finally, these two benefit-per-ton estimates may diverge due to inherent differences in the model used to estimate air quality impacts. As described above, EPA used a dispersion model to estimate diesel PM air quality changes; conversely, EPA used a photochemical grid model to generate air quality estimates for the benefit-per-ton estimates that supported the ozone RIA.

Table 2: Counties with Highest Predicted Benefit-per-ton Estimates ($/ton) for Total Diesel Sources.

Table 3:  Counties with Highest Predicted Benefit-per-ton Estimates ($/ton) for On-road Diesel Sources.

Table 4:  Counties with Highest Predicted Benefit-per-ton Estimates ($/ton) for Non-road Diesel Sources.

Table 5:  Benefit-per-ton of Diesel Emissions Reduced ($/ton) for Counties with the Lowest Emissions of Total Diesel Sources.

Table 6:  Benefit-per-ton of Diesel Emissions Reduced ($/ton) Results for Counties with the Lowest Emissions of On-road Diesel Sources.

Table 7: Benefit-per-ton of Diesel Emissions Reduced ($/ton) Results for Counties with the Lowest Emissions of Non-road Diesel Sources.

Table 8:  Counties with Highest Import/Export Factors for Total Diesel Sources