Valuing the Ozone-Related Health Benefits of Methane Emission Controls

2.1 Methane Emissions and Ozone Concentrations

2.2 Ozone-Mortality Concentration-Response Relationships

We use epidemiologically-derived health impact functions to estimate changes in both premature short-term respiratory mortality related to transient changes in ozone concentration and longer-term mortality from chronic exposure to peak ozone concentrations in response to a pulse of methane emissions.

We quantify ozone-mortality impacts using two different ozone exposure metrics found by the epidemiological literature to be associated with premature mortality: daily exposure (“short-term”) and repeated exposure to hourly peak concentrations over the warm season (“long-term”). For short-term exposure, we calculate \(\upbeta \) from the (Bell et al. 2004) study of 95 US urban communities, who found that a 10 ppb increase in 24-h average ozone was associated with a 0.64 % (95 % posterior interval, 0.31–0.98 %) increase in cardiovascular and respiratory mortality. For long-term mortality, we calculate \(\upbeta \) from the (Jerrett et al. 2009) study of the American Cancer Society study cohort of adults aged 30 and older. We use the two pollutant model that controlled for fine particulate matter (PM2.5), which found that a 10 ppb increase in the April-September average of the 1-h daily maximum ozone concentration was associated with a 4 % (95 % confidence interval, 1.0–6.7 %) increase in respiratory mortality. Following (Anenberg et al. 2010, 2012), we apply these concentration-response relationships found in the US globally, despite differences in exposure, susceptibility, and vulnerability around the world. We focus on cause-specific mortality since it is more comparable among populations in different countries than is all-cause mortality.

2.3 Future Population Projections

To estimate the changes in mortality over time due to a reduction in \(\hbox {CH}_{4}\), it is necessary to project population (Eq. 3, pop) and baseline mortality (Eq. 3, \(\hbox {y}_{0}\)) over the coming century. While a number of long-term projections for global population exist, e.g. (Nakicenovic 2000; United Nations 2004), in order to promote consistency and comparability with the SCC approach, the same scenarios used by the US Government in developing and updating the SCC (US Department of Energy 2010; US Government 2013) are used here.⁵ These scenarios are Business as Usual (BAU) simulations from four integrated assessment models—MiniCAM, MESSAGE, MERGE, and IMAGE—that participated in the Energy Modeling Forum 22 (EMF 22) exercise (Clarke et al. 2009). The population for each of the four models’ scenarios is aggregated to eight global regions as described in (US Department of Energy 2010). Our base estimate uses the MiniCAM scenario and the other three scenarios are used to test the sensitivity of the results to changes in population and GDP.

We estimate the effects of long-term exposure to peak ozone concentrations only for the population ages 30 and above, consistent with the American Cancer Society cohort used by Jerrett et al. (2009). In order to generate this value for each of the four socioeconomic scenarios, we first estimated the fraction of the population in this demographic using UN country-level population projections (United Nations 2004) by age category and then aggregated to our eight regions. We then applied this ratio to the total population for each of our four socio-economic scenarios.⁶

2.4 Future Baseline Mortality Rates

Our methodology to estimate mortality due to ozone concentrations changes also requires baseline rates of death due to cardiovascular and respiratory disease for the total population and respiratory disease for the population age 30 and above.

These rates vary greatly by region and would be expected to change over time as economies develop and mortality rates shift from infectious disease to chronic disease. This type of information is not included in the EMF 22 scenarios used for the SCC estimation. Country-level projections of total cardiovascular and pulmonary deaths through 2100⁷ [International Futures Project (IFs) base projections (International Futures 2013)] are aggregated to the eight regions corresponding to the population and GDP scenarios, weighted by 2004 country population and cardiovascular and pulmonary death rates (Mathers et al. 2008). We apply these baseline death rates to all socio-economic scenarios. We test the sensitivity of the results to baseline mortality rates using constant 2010 baseline mortality rates.

Estimating long-term ozone mortality impacts requires baseline respiratory mortality rates for the subset of the population ages 30 and above. We first use the Global Burden of Disease (Lim et al. 2012)⁸ country-level, ages 30+, respiratory mortality rates for 2010, weighted and aggregated to the eight regions, then approximate annual rates through 2100 by scaling 2010 respiratory mortality with the cardiovascular and pulmonary mortality rates from the IFs. We also test the sensitivity of results to constant 2010 mortality rates.

2.5 Valuation of Changes in Mortality

In order to monetize the reduction in mortality estimated above, we apply a Value of Statistical Life (VSL) to the change in mortality in each future year. These values are then discounted back to the year of emission and summed, producing a Net Present Value (NPV) of the indirect health benefits of \(\hbox {CH}_{4}\) mitigation, measured in (2011)$/t–\(\hbox {CH}_{4}\).

The total annual value of reduced mortality must be discounted back to the year of emissions in order to calculate the NPV. Estimates of benefits that accrue over long time horizons (such as the SCC) are highly sensitive to the discount rate. We therefore test seven discount rate estimates. The first four are constant discount rates of 2.5, 3.0, 5.0, and 7.0 %. The first three correspond to the discount rates used to develop the US SCC estimates and 3 and 7 % correspond to the discount rates of consumption and capital as used in US regulatory analyses (US Office of Management and Budget 2003).

The base estimates for both short- and long-term ozone-related mortality due to a 1 tonne change in methane are calculated using an emission year of 2020, the MiniCAM scenario inputs for population and GDP, average \(\Delta \hbox {O}_{3}\), central estimates of \(\upbeta \), the IFs baseline for \(\hbox {y}_{0}\), and a constant discount rate of 3 %. We test the sensitivity of the ozone-related mortality benefits of a reduction of a marginal tonne of methane over all input parameters. These include four additional estimates of the attributable fraction of mortality from methane emissions, including both a high and low estimate of \(\Delta \hbox {O}_{3}\) and the bounds of the 95 percentile range estimate of \(\upbeta \), an alternate specification of \(\hbox {y}_{0}\), GDP and population from three additional socioeconomic scenarios, six additional discount rates, two additional methods of VSL estimation, and four other years of emission.

3.1 Sensitivity Analyses

We tested the sensitivity of the results to a number of parameter assumptions and explored the relative importance of different factors in determining the mortality benefit, as described above. These sensitivities were all calculated relative to the base case. Sensitivity ranges were chosen based on ranges for parameters provided within underlying studies (\(\Delta \hbox {O}_{3},\,\upbeta \), Ramsey discount rates), sensitivities used within the US Government SCC approach (discount rates 2.5, 3, and 5 %, different scenarios, emission years), OMB Circular A-4 (discount rate of 7), and reasonable bounding parameters chosen by the authors (VSL elasticities and CVP mortality rate). Therefore, while these calculations are informative, this sensitivity analysis is not comprehensive nor are the ranges as comparable as they would be if they were all equivalent standard deviation ranges.

Figure 1 shows the difference between the base estimate, (2011)$790/t–\(\hbox {CH}_{4}\), and each sensitivity and Table 1 reports the estimates for each of the alternate measures for the specified parameters. The estimates are very sensitive to the elasticity parameter used to extrapolate the VSL over region and time. The use of a constant VSL \((\upvarepsilon = 0)\) nearly doubles the base estimate in 2020, while the VSL estimated as a strict ratio of per capita income \((\upvarepsilon = 1)\) produces an estimate only half of the base value.

Open image in new window

Figure 2 shows the effect of the different VSL metrics on the \(\hbox {CH}_{4}\) mortality benefit over time. The estimate using \(\upvarepsilon = 1\) is one-half the base estimate in 2020 and increases to two-thirds the base estimate in 2050. Similarly, an elasticity of 0 produces a mortality benefit roughly twice as large as the base estimate in 2020, but only 1.5 times larger in 2050. That the estimates using a constant VSL \((\upvarepsilon = 0)\) are much larger than with the base method is not particularly surprising, as the US 2010 per capita income is substantially greater than most regions in the world at present and greater than what some regions reach even by the end of the century in these scenarios.

The year of emission also has a large effect on the estimate, with benefits per tonne more than doubling for methane emitted between 2010 and 2050 (Fig. 2). Within the base set of parameters, this change is due primarily to increases in population and VSL changes due to income growth over the time period.

We test a wide range of discount rates and methods and find that the estimate of the indirect short-term mortality impacts from a tonne of methane emitted in 2020 using a constant 2.5 % discount rate is about 1.3 times the estimate with a constant 5 % discount rate. In contrast, the SCC is very sensitive to the choice of discount rate and methodology; the SCC is nearly 5.5 times greater when discounted at 2.5 versus 5 %. However, Waldhoff et al. (2014) show that direct estimates of the SC–\(\hbox {CH}_{4}\) are also much less sensitive to discounting than the SCC, due to methane’s comparatively short atmospheric lifetime and the resulting nearer-term nature of the damages. The ozone-mediated mortality benefits from methane emissions are even less sensitive to discounting. This is likely due to two reasons. First, the inertia of the climate system will cause temperature changes that last longer than concentration changes and therefore the climate damages, but not the ozone-mediated mortality damages, will continue to increase with future temperature changes. Second, the social cost of methane is a function of the reference temperature as well as population and GDP. Because temperature is increasing in the reference scenarios (even without the addition of a marginal tonne of \(\hbox {CH}_{4}\)), the climate damages from future climate changes are larger than they would be under present temperature conditions so the climate damages decay slightly more slowly than the mortality damages that are not dependent on temperature.

Open image in new window

We examine four different future socio-economic scenarios, namely reference scenarios from the EMF 22 exercise from the IMAGE, MERGE, MESSAGE, and MiniCAM models. The use of different socioeconomic scenarios can increase the base estimate from up to 6 % in 2010 to up to 12 % for emissions in 2050. The increasing sensitivity across scenarios over time is a result of the growing differences across the models’ projections of GDP and population as the scenarios extend into the latter part of this century.

We also examine the sensitivity of incidence results to year of emission, CVP mortality response, \(\hbox {O}_{3}\) concentration-response factor, CVP mortality rates, and socioeconomic scenario (Figure 3).

Open image in new window

The concentration-response factor \((\upbeta )\) and the year of emission reduction have the largest effect on avoided premature mortality. The \(95\mathrm{th}\) percentile ranges of \(\upbeta \) bound the avoided premature mortality between 116 and 366. Over time, a reduction of 1 MMT–\(\hbox {CH}_{4}\) in a single year increases the total avoided deaths, from 210 in 2010 to 329 in 2050.

To put these values in perspective, the recent EPA Oil and Natural Gas Sector: New Source Performance Standards and National Emission Standards for Hazardous Air Pollutants Reviews rule¹⁷ estimated that actions to reduce VOC emissions from oil and gas production would also reduce methane emissions by 1.0 million tonnes annually, with an additional 0.7 million tonnes attributed to voluntary reductions. The net effect of these reduced \(\hbox {CH}_{4}\) emissions in the year 2020 is to avoid 406 premature deaths globally from the resulting reduction in ozone concentrations over the following 70 years. The net present value of the avoided deaths due to reductions in 2020 \(\hbox {CH}_{4}\) emissions is $1.3 billion using our base estimate.

3.2 Regional Estimates of Cardiovascular and Pulmonary Mortality Damages

In order to examine the distributional impacts of methane on ozone health, we present estimates of avoided mortality and benefits by region. As seen in Fig. 4, the estimates vary greatly across regions, ranging from a low of $44/t–\(\hbox {CH}_{4}\) in Africa and the Middle East to a high of $181/t–\(\hbox {CH}_{4}\) in East Asia under base assumptions. That the US falls somewhat in the middle of this range may seem somewhat surprising. However, the estimates are dependent not only on the VSL within a region, but also on the total population and the baseline cardiovascular and pulmonary death rates. Large populations and high cardiovascular and pulmonary mortality may cause the total benefits to exceed those in regions with much higher per capita incomes. This is demonstrated in Table 2: a one million tonne decrease in emissions in 2020 avoids 12 deaths in the United States, compared to 68 avoided deaths in East Asia. The effects of per capita income (the basis for the VSL estimates) are worth noting, however. While only 4.5 % of the world’s population lives in the United States, 10.6 % of the benefits are accrued there, primarily due to differences in the estimated VSL.

Table 2

Total avoided mortality per 1 MMT decrease in \(\hbox {CH}_{4}\) in a given year (base case)

Year	2010	2020	2030	2040	2050
European Union	26	29	30	32	31
United States	10	12	14	16	17
Other OECD	8	10	12	13	14
Eastern Europe	21	23	25	27	29
East Asia	56	68	81	93	102
Southeast Asia	57	64	73	83	93
Africa-Middle East	19	18	18	18	19
Latin America	13	16	18	21	24
Global	210	239	271	302	329

Due to rounding, global totals may not equal the sums of regional values

Open image in new window

As with the global values, the benefits accrued by individual regions is very sensitive to the VSL methodology, though the magnitude, and even sign, varies across regions. When VSL is a constant value \((\upvarepsilon = 0)\), the US value is roughly 5 % of the global estimate. In sharp contrast, a VSL estimate that varies linearly with per capita income \((\upvarepsilon = 1)\) increases the fraction of benefits that accrue in the US to 25 %. The effect is opposite to the effects of VSL methodology on all other regions, where the use of a constant \((\upvarepsilon = 0)\) VSL increases benefits and the linear method \((\upvarepsilon = 1)\)decreases benefits compared to the base assumption of \(\upvarepsilon = 0.4\). This is because the US has a relatively high VSL compared with other regions, particularly in the near-term. A constant 2010 US VSL raises other regions’ VSLs by more than it decreases the US value over time. When \(\upvarepsilon = 1\), however, the US VSL increases linearly, rather than exponentially at 0.4. Since other regions have lower incomes, the exponent \((\upvarepsilon = 0.4)\) increases VSL relatively more, and future increases in income in the US do not change the VSL as much as with a \(\upvarepsilon = 1\) method.

Further differences can be seen across the magnitude of response to VSL methodology by region (Fig. 4). The use of a constant VSL across region and time has the largest effect on the Africa-Middle East and Southeast Asia regions, increasing the benefits in each region nearly 175 % compared to the base estimate. Generally, regions with relatively low per capita incomes have much larger increases in benefits when \(\upvarepsilon = 0\) than corresponding decrease when \(\upvarepsilon = 1\). Higher income regions (e.g. European Union and Other OECD) have comparable changes between the two methods.

For further comparison, Table 2 shows the total avoided deaths for a one million tonne decrease in methane emissions in each year. While economic growth, across regions and over time, has a large effect on the monetized benefits, the drivers of avoided mortality are population and baseline mortality rates, both of which vary by region and over time.

Total avoided deaths for 1 MMT–\(\hbox {CH}_{4}\) emissions reduction increases over time most rapidly for three regions: Other OECD, East Asia, and Latin America, all seeing avoided mortalities increase by more than 80 % between 2010 and 2050. These increases are driven primarily by rapidly increasing cardiovascular and pulmonary death rates. Similarly, the Africa and Middle East region shows practically no change in total avoided mortality over time. This is because the increase in population in this region is almost perfectly offset by the decrease in cardiovascular and pulmonary death rates through mid-century.¹⁸ This region begins the century with the lowest cardiovascular and pulmonary death rates, and though these rates are projected to begin increasing again around 2050, by century’s end the rate is still lower than in 2010.

3.3 Long-Term Ozone Mortality and Damages

In addition to cardiovascular and pulmonary mortality from short-term exposure to peak ozone concentrations, changes in methane emissions can also impact respiratory mortality from long-term exposure to average ozone concentrations. We examine the mortality effects and damages of chronic ozone exposure due to changes in methane emissions and test the sensitivity of the results to several parameters. While the long-term mortality results respond similarly to the short-term estimates for most of the sensitivity tests, there are two differences worth noting. The long-term concentration-response factor has a relatively large 95 % confidence interval. Because of this, the long-term respiratory mortality results are very sensitive to this parameter, with a range of (2011)$444–$2973.

Additionally, as seen in Fig. 5, the benefits of avoided respiratory morality due to long-term ozone exposure increase over time at a faster rate than for the avoided cardiovascular and pulmonary mortality due to short-term ozone exposure. The long-term benefits per tonne of \(\hbox {CH}_{4}\) mitigated increase from $1301 in 2010 to $3846 in 2050, while the short-term benefits increase from $634 to $1,428 in the same period. The primary reason for the faster increase in benefits is that the long-term mortality impacts are based only on the sub-set of the population ages 30 and above. Globally, the affected demographic (population 30+) for the long-term impacts is growing faster than the total population, upon which the short-term estimates are based.

Open image in new window

The long-term mortality relationships include both acute and chronic effects of ozone exposure, so the significantly larger long-term mortality estimates suggest that considering only short-term mortality may exclude a substantial portion of ozone-related risk. However, because the short-term mortality relationships include a larger population (all ages versus adults ages 30 and older only) and mortality due to any cardiovascular and pulmonary disease rather than just respiratory disease, the short-term mortality estimates may include some ozone mortality effects that are not captured in the long-term results. As the extent of the overlap between these estimates is unknown, the short-term and long-term mortality results should be viewed as complementary estimates, rather than additive or one as a subset of the other.

These results can be compared to work by (Shindell et al. 2012) that shows long-term global methane health benefit estimates in the range of $1080 \(\pm \) $721/t–\(\hbox {CH}_{4}\) (in 2006$).¹⁹ There are several of differences between the Shindell et al. approach and the methods used in this paper. The key difference is that the estimates presented in this paper follow the decay of a pulse of methane over a period of time and reflect the value of the change in mortality due to the resulting ozone concentration to the population in the same time period. For a pulse of methane emitted in 2020, the ozone increment in 2020 is applied to the population in 2020, the ozone increment in 2021 is applied to the population in 2021, and so forth, and the changes in mortality in each region and year are valued with a region and time consistent with estimated VSL. In contrast, Shindell et al. used a more sophisticated chemistry-climate model to calculate the ozone changes driven by emission changes in 2030 and averaging the ozone response in years 30–50 of the simulation. This ozone increment is then applied to the population of 2030 in order to estimate changes in mortality. The Shindell et al. approach would be expected to yield a result more similar to the integrated valuation of a single year’s emissions (as in this paper) with no future changes in population, GDP, or mortality response, and if future benefits were not discounted. In addition, Shindell et al. include \(\hbox {PM}_{2.5}\) concentration changes and health impacts resulting from \(\hbox {CH}_{4}\) emission reductions, whereas we focus here on ozone only.

3.4 Caveats and Limitations

A few of the assumptions made to develop these benefits estimates have the potential to slightly increase the range of possible outcomes. For simplicity these estimates assume that the ozone concentration response to methane is homogenous. However, (Anenberg et al. 2012) suggest accounting for the heterogeneity in ozone response to \(\hbox {CH}_{4}\) may lead to differences in mortality estimation on the order of 15 % globally, with uncertain sign. The decay rate of methane is also not a constant, with some papers suggesting that it might change by up to 10 % depending on emissions of VOCs, \(\hbox {NO}_\mathrm{X}\), and methane itself (Sarofim 2012). A change in the overall decay rate would presumably also have an effect on the rate of ozone production. However, these effects are relatively small compared to the uncertainties involved in the choices of VSL and discount rates considered above.

Due to lack of regional studies, we also apply concentration-response functions found in the US globally, though differences in exposure and population susceptibility characteristics (e.g. time spent outdoors, home ventilation rates, medical care, etc.) could cause differences in this factor across regions. As discussed in Anenberg et al. (2010), short-term \(\hbox {O}_{3}\) epidemiology studies in developing nations produce fairly similar results to short-term studies in North America and Europe (Health Effects Institute International Scientific Oversight Committee 2004). Furthermore, concentration–mortality relationships do not vary significantly by sex, age, and race (Jerrett et al. 2009; Krewski et al. 2009; Zanobetti et al. 2000), although some sensitive populations may be at a higher risk. We base our estimates on cardiovascular and respiratory mortality rates rather than all-cause mortality rates because this will reduce errors due to differences in causes of deaths across regions.

These estimates also do not include the benefits of avoided morbidity. The additional value of reduced morbidity from reductions in ozone concentrations due to lower methane emissions would likely increase the total health benefit valuation, though precisely how much is left to future work. Additional effects that would serve as a complement to both the climate effects from the social cost of methane and the health effects from this analysis include the effect of ozone concentrations on agricultural and forestry yields (Avnery et al. 2013) and on the carbon uptake of natural systems (Feltzer et al. 2005).

One effect that could lead to our results being overestimates is our exclusion of particulate matter concentrations and health impacts. Reduced methane emissions may affect atmospheric oxidant concentrations, leading to increased production of sulfate, a component of particulate matter pollution (Anenberg et al. 2012). Increased particulate matter-related health impacts resulting from reduced methane emissions may counteract some of the ozone-related health benefits.