Appendix 6 Methods for the Analysis of the Benefits of Abatement

Overview

This Appendix identifies the data sources used for valuation, exposure-response and other data.

The overall structure of the analysis proceeds through the series of stages shown in Figure 2 of Chapter 5. Emissions are modelled to provide data on the change in concentration and deposition across the UK (using the HARM and ELMO models) and Europe as a whole (using the EMEP models). The dispersion/concentration maps are then overlaid onto data showing the distribution of the ‘stock at risk’ (people, buildings, crops, etc.), and information on the sensitivity of the stock at risk (death rates, age structure of the population, type of ecosystems, etc.). This provides information on exposure, to which can be applied exposure-response functions to derive estimates of impact in biological and physical terms. The final stage is valuation of impacts, where possible.

As stated elsewhere in this report, there are significant uncertainties in the assessment of the benefits of air pollution abatement. However, we believe that it is possible to take an objective view in the analysis through comprehensive treatment of these uncertainties.

The analysis generated several types of result:

1. UK benefits from UK abatement (‘Partial-UK’)
2. UNECE benefits from UK abatement (‘UNECE’)
3. UK benefits from pan-UNECE abatement (‘Full-UK’)

Group i effects

Materials damage

UK assessment: IGCB report (DETR, 1998)

UNECE assessment: Holland et al (1999)

Crops damage

UK assessment: IGCB report (DETR, 1998)

UNECE assessment: Holland et al (1999)

Ozone exposures

RAINS model, web version (see Appendix 5)

Acidification and eutrophication

ITE databases (Full-UK, see also Appendix 7) and RAINS model, web version (Partial-UK and UNECE, see also Appendix 5).

Particulate exposures

HARM, ASAM for the full-UK analysis.

EMEP for the Partial-UK and UNECE analysis.

Acute effects on mortality and respiratory hospital admissions (RHAs)

Exposure response functions from COMEAP, including 95% confidence limits for error analysis.

Group II Effects

Numerous acute effects on morbidity

Functions and ranges from COMEAP and EAHEAP for the following effects:

1. Cardiovascular disease
2. Congestive heart failure
3. Ischaemic heart disease
4. Cerebrovascular hospital admissions
5. Bronchodilator usage for asthmatic adults and children
6. Cough for asthmatic adults and children
7. Wheeze for asthmatic adults and children.

Group III effects

Monetisation of acute effects on mortality and RHAs

Acute mortality was valued using the EAHEAP range. Lower bound estimates of cases (as calculated above) were combined with the lower bound from EAHEAP (£2,600), and vice versa for the upper bound (valued at £1.4 million/case). An intermediate case taking the COMEAP mid-point and EAHEAP intermediate valuation of £110,000/case was also calculated, but not used in the accumulated ranges.

Group IV effects

Monetisation of various acute effects on morbidity (from GROUP II)

All effects valued using ExternE data (European Commission, 1995, 1999). See also the note on Group II effects, describing the way in which data were combined to give the accumulated high and low estimates of benefit, to avoid possible double counting.

Group V Effects

Chronic mortality

This part of the analysis was based on work carried out by Fintan Hurley and Brian Miller at the Institute of Occupational Medicine, using the Pope et al function. The underlying analysis is complex, being based on manipulation of life-tables. The work is summarised in Markandya et al (1999) and a forthcoming report by Hurley et al for the Department of Health. Numerous assumptions are available for the assessment of chronic mortality impacts. For this analysis it was assumed that the change in risk was the same at all ages, with effects calculated for (essentially) a one-year reduction in particle concentrations. Other assumptions are plausible and will be explored in the Hurley et al report.

Unlike the estimates of acute effects on mortality expressed in terms of number of cases (with no data on the reduction in life expectancy), the estimated chronic effects on mortality are quantified in terms of life years lost (with no data on the number of cases). As the EAHEAP low (£2,600/case with an average of 1 month of life lost) and intermediate (£110,000/case with an average of 1 year of life lost) points provide an estimate of willingness to pay for increased longevity, we have applied these data. There will be debate on the applicability of these valuations for chronic mortality. However, it would appear to us to be curious if the final valuations agreed were so substantially different that they changed the results gained from applying this broad range.

Group VI Effects

Forests, numerous chronic and acute effects on morbidity

There is significant debate on the meaning of the effects listed in this Group. In the UK, for example, chronic bronchitis is not an illness associated with children, though the US literature assesses it as an endpoint. For chronic bronchitis in adults it would appear from the forthcoming report by Hurley et al (1999) that there are problems in distinguishing between those contracting bronchitis for the first time ever, and those suffering repeat episodes. Similarly there are difficulties in interpretation of ‘restricted activity days’ (RADS).

In view of these problems with interpretation we have applied broad ranges for the valuation of the most prominent of this group of effects. The main data sources are Holland et al (1999), Hurley et al (1999, forthcoming) and European Commission (1995, 1999).

The tables in Appendix 8 provide running totals for benefits, starting with those effects that can be quantified with the best confidence and proceeding through benefits of diminishing certainty. This is done in such a way as to provide both lower and upper bounds. High and low values are selected for inputs where required in order to derive these ranges.

Data Used in the Analysis

Modelling Pollution Concentrations and Deposition

Most of the pollution dispersion and chemistry analysis carried out for this study was based on the use of the HARM and ELMO models and was described in Appendix IV. This analysis was carried out to give a resolution of 10 x 10 km. Analysis for the UNECE area was based on the EMEP 150 x 150 km grid. For the UNECE scale, model runs were carried out at the Norwegian Meteorological Institute, relating emissions in a number of years to air concentrations and deposition of all of the pollutants relevant to this study. Average transfer coefficients for each country to grid cell combination were calculated for a six year period (covering 1989 to 1994), in order to account for meteorological variation. Data on ozone were supplied specifically for several of the scenarios assessed under separate analyses for the UNECE Task Force on Economic Aspects of Abatement Strategies, based on average meteorology over a five-year period.

Health Effects Assessment

Stock at risk data and atmospheric modelling

For the UK, population data were taken from the 1991 census, disaggregated by age, and expressed on a 10 x 10 km grid to match the pollution data.

The main source of population data used for the UNECE area analysis was the RIVM land use database (Veldkamp and van der Velde, 1995). These data were transferred to the EMEP 150 x 150 grid, and disaggregated to urban and rural populations. The Bosnian population, absent from the original inventory, was taken from UN sources.

Additional data were required to define the fraction of the population in various groups considered to be at special risk - the elderly, children, and asthmatics. In addition, death rate data were required. Data on age structure and death rates were obtained from Rayner et al (1994), drawing on relevant UN reports (Demographic Yearbook, Population and Vital Statistics Report and World Population Prospects). Over Europe these provide average factors as follows;

Fraction of children in European population: 0.2

Fraction of adults in European population: 0.8

Fraction of people > 65 years in European population: 0.14

Annual death rate per thousand people: 10.2

The following estimates are made for asthmatics (R. Anderson, personal communication, October 1997);

Child asthmatics as a fraction of the UK population: 0.02

Adult asthmatics as a fraction of the UK population: 0.04

Exposure-response functions

Available literature on health effects of air pollution has been reviewed by the Department of Health (1998) and by Hurley, Donnan and their colleagues at IOM, providing the exposure-response functions listed in Table AII.1. Preference is given to the DH preferred function where both it and supporting data on incidence of effects in the UK are available.

Table A6.1. Quantification of human health impacts. The exposure response slope, f_er, has dimensions of [units/(yr-person-ug/m³)] for morbidity excluding RHAs, [% change in annual mortality,RHA rate/(ug/m³)] for acute effects on mortality ad RHAs, and years of life lost for chronic effects on mortality.

Receptor

Impact Category

Reference

Pollutant

fer 1

ASTHMATICS

adults	Bronchodilator usage	Dusseldorp et al, 1995	PM10	0.163
	Cough	Dusseldorp et al, 1995	PM10	0.168
	Lower respiratory symptoms (wheeze)	Dusseldorp et al, 1995	PM10	0.061
children	Bronchodilator usage	Roemer et al, 1993	PM10	0.078
	Cough	Pope, Dockery, 1992	PM10	0.133
	Lower respiratory symptoms (wheeze)	Roemer et al, 1993	PM10	0.103
all	Asthma attacks (AA)	Whittemore, Korn, 1980	O3	4.29E-3

ELDERLY 65 years +

Congestive heart failure (CHF)

Schwartz, Morris, 1995

PM10

1.85E-5

CHILDREN

	Chronic bronchitis	Dockery et al, 1989	PM10	1.61E-3
	Chronic cough	Dockery et al, 1989	PM10	2.07E-3

ADULTS

	Restricted activity days (RAD)3	Ostro, 1987	PM10	0.025
	Minor restricted activity day(MRAD)4	Ostro, Rothschild, 1989	O3	9.76E-3
	Chronic bronchitis	Abbey et al, 1995	PM10	4.9E-5

ENTIRE POPULATION

	Respiratory hospital admissions (RHA)	EAHEAP (1999)	PM10 SO2 O3	0.08% 0.05% 0.07%
	Cerebrovascular hospital admissions (CVA)	Wordley et al, 1997	PM10	5.04E-6
	Symptom days	Krupnick et al, 1990	O3	0.033

DEATH RATES

	Acute Mortality	EAHEAP (1999)	PM10 SO2 O3	0.075%0.06% 0.06%
	Chronic Mortality	Pope et al, 1995	PM10	0.00036

¹ Sources: [ExternE, European Commission, 1995; 1999] and [Hurley and Donnan, 1997].

³ Assume that all days in hospital for RHA, CHF and CVA are also restricted activity days (RAD). Also assume that the average stay for each is 10, 7 and 45 days respectively.

Thus, net RAD = RAD - (RHA*10) - (CHF*7) - (CVA*45).

⁴ Assume asthma attacks are also MRAD, and hence should deducted from the MRAD total.

Use of the non-COMEAP functions listed here is based on the assumption that the incidence rates of different types of illness are the same in the UK and elsewhere in Europe as in the location where studies were originally undertaken. The calculations required were carried out by Hurley and colleagues at IOM in the EC DGXII ExternE Project. The following illustrative case was kindly supplied for this project by Fintan Hurley. The Table numbers referred to are to be found in the paper by Roemer et al (1993).

Calculations on bronchodilator usage, from Roemer et al. (1993)

1. Prevalence from Table 4. Take a weighted average of 9.0 (27 days); 10.3 (32d); 11.5 (11d) and 12.1 (9d) to get 10.2.
2. Regression coefficient from Table 6. Treated as if it is linear regression; though it may be logistic. The answers should be similar. Taking it as linear, we have:
3. Change in prevalence per mg.m^-3 PM₁₀ is 0.023% or 0.23% per 10 mg.m^-3 PM₁₀.
4. With a s.e. of 0.008, the 95% CI is given by ± 2s.e.; i.e. ±0.16%; so for 10 mg.m^-3 PM₁₀ we have a 95% CI of 0.07%, 0.39%.
5. Sticking with the main calculation, the average prevalence is 10.3%; and so, for an extra 10 mg.m^-3 PM₁₀, we get a prevalence of 10.53%.
6. The % increase in prevalence is 0.23/10.3 = 0.023, i.e. a RR of 1.023 per 10 mg.m^-3 PM₁₀.
7. Still taking it as linear, we have:

Per 1 mg.m^-3 PM₁₀, an increase of 0.023% prevalence on a given day; i.e.

0.00023 x 365 = 0.084 occurrences per child asthmatic per year; i.e.

84 extra occurrences per 1000 child asthmatics per year, where ‘occurrence’ is the use of a bronchodilator on one day where otherwise there would have been no usage.

We used 77.9 rather than 84, because we applied the increase as a RR to the odds, as in logistic regression, not as a linear additive effect, as above.

The assessment of the chronic effects on mortality was based on a reasonably central series of model runs carried out at IOM using life tables and described in a separate study (Markandya et al, 1999). The analysis assessed the effect of a one year pulse of pollution on the life expectancy of the UK. It was assumed that the change in mortality risk was the same for all people exposed. This analysis, and the effect of alternative assumptions will be described more fully in a forthcoming report to Department of Health (Hurley et al, 1999).

The ozone assessment was carried out assuming the presence of a threshold of 50 ppb only, following from the EAHEAP analysis. This should not be taken as presumption against the no-threshold position, it was done simply because of a lack of modelled data for the UK to show trends in mean exposure to ozone for the scenarios considered. The threshold calculations were carried out in a rather indirect manner, assuming that exposure to levels in excess of 50 ppb will be reflected in the change in AOT40 and AOT60. Appropriate ozone data were limited to 1990 and 2010. The assessment proceeded through the following stages:

1. It was assumed that the results presented by EAHEAP were for a year in which average ozone was similar to 1990 concentrations.
2. The EAHEAP ozone results assuming a 50 ppb threshold were divided by population-weighted AOT60 for 1990 to give deaths/RHAs per million person ppm.hrs.
3. The result of [2] was multiplied by the change in AOT60 million person ppm.hrs to give the change in deaths and RHAs per scenario.
4. Stages [2] and [3] were repeated using the AOT40.
5. An average was taken of the results of [3] and [4].

Endpoint	Value	Estimation Method and Comments
Acute Morbidity
Restricted Activity Day (RAD)	63	CVM in US estimating WTP.
Symptom Day (SD) and Minor Restricted Activity Day	6.3	CVM in US estimating WTP. Account has been taken of Navrud’s study.
Chest Discomfort Day or Acute Effect in Asthmatics (Wheeze)	6.3	CVM in US estimating WTP. Same value applies to children and adults.
Emergency Room Visits (ERV)	186	CVM in US estimating WTP.
Cardiovascular Hospital Admissions	6,560	As above.
Acute Asthma Attack	31	COI (adjusted to allow for difference between COI and WTP). Applies to both children and adults.
Chronic Morbidity
Chronic Illness (VSC)	1,000,000	CVM in US estimating WTP.
Chronic Bronchitis in Adults	88,000	Rowe et al (1995).
Non fatal Cancer	375,000	US study.
Malignant Neoplasms	375,000	Valued as non-fatal cancer.
Chronic Case of Asthma	88,000	Based on treating chronic asthma as new cases of chronic bronchitis.
Cases of change in prevalence of bronchitis in children	225	Treated as cases of acute bronchitis.
Cases of change in prevalence of cough in children	188	As above.

Damage to Materials

Stock at risk data

The stock at risk is derived from data on building numbers and construction materials taken from building survey information. Sources of data are as follows;

Eastern Europe (including the former East Germany):

Kucera et al (1993b), Tolstoy et al (1990) - data for Prague

Scandinavia:

Kucera et al (1993b), Tolstoy et al (1990) - data for Stockholm and Sarpsborg

UK, Ireland:

Ecotec (1996), data for UK extrapolated to Ireland

Former West Germany:

Hoos et al (1987) - data for Dortmund and Köln

Other western Europe:

Average of material use per person from Hoos et al, Kucera et al and Tolstoy et al (excluding Prague), and Ecotec.

For galvanised steel in structural (non-building) applications an average was derived from European Commission (1995) and Kucera et al (1993b).

Meteorological, atmospheric and background pollution data

The exposure-response functions require data on meteorological conditions. Of these, the most important are precipitation and humidity. Data have been taken from Kucera (1994).

Dose-response functions

The main source of data for exposure response functions used here is the work conducted under the UNECE Programme (Kucera, 1993a, 1993b, 1994). This section lists the dose-response functions used, which should be assumed to originate from the work of Kucera unless otherwise referenced. The following key applies to all equations given:

ER = erosion rate (um/year)

P = precipitation rate (m/year)

SO₂ = sulphur dioxide concentration (ug/m³)

O₃ = ozone concentration (ug/m³)

H⁺ = acidity (meq/m²/year)

R_H = average relative humidity, %

f₁ = 1-exp[-0.121.R_H/(100-R_H)]

TOW = fraction of time relative humidity exceeds 80% and temperature >0·C

ML = mass loss (g/m²) after 4 years

In all the ICP functions, the original H⁺ concentration term (in mg/l) has been replaced by an acidity term using the conversion:

P^.H⁺ (mg/l) = 0.001^.H⁺ (acidity in meq/m²/year)

To convert mass loss for stone and zinc into an erosion rate in terms of material thickness, respective densities of 2.0 and 7.14 tonnes/m³ are assumed. The functions used are as follows;

Unsheltered limestone (4 years): ML = 8.6 + 1.49.TOW.SO₂ + 0.097.H⁺

Unsheltered sandstone (4 years) (also mortar): ML = 7.3 + 1.56.TOW.SO₂ + 0.12.H⁺

Brickwork: no effect

Concrete; assumed no effect, though air pollution may affect steel reinforcement

Carbonate paint: DER/tc = 0.01[P]8.7(10^-pH - 10^-5.2)+0.006.SO₂.f₁ (Haynie, 1986)

Silicate paint: DER/tc = 0.01[P]1.35(10^-pH - 10^-5.2)+0.00097.SO₂.f₁ (Haynie, 1986)

Steel: assumed either painted or galvanised, not assessed independently

Unsheltered zinc (4 years): ML = 14.5 + 0.043.TOW.SO₂.O₃ + 0.08.H⁺

Sheltered zinc (4 years): ML = 5.5 + 0.013.TOW.SO₂.O₃

Aluminium: assumed too corrosion resistant to be affected significantly.

Calculation of repair frequency

It is assumed that maintenance is ideally carried out after a given thickness of material has been lost. This parameter is set to a level beyond which basic or routine repair schemes may be insufficient, and more expensive remedial action would be needed. A summary of the critical thickness loss for maintenance and repair are shown in Table AII.6.

Table A6.3. Assumed critical thickness for maintenance or repair measures for building materials.

Material	Critical thickness loss
Natural stone	5 mm
Rendering	5 mm
Mortar	5 mm
Zinc: Construction - sheet and strip Other construction, agriculture and street furniture Pylons, other transport	25 um 50 um 100 um
Galvanised steel	50 um
Paint	20 um

Repair costs

The following repair costs were applied for the interval in material lifetime between scenarios defined by the application of data on concentration, dose-response and critical thickness.

Table A6.4. Repair and maintenance costs [euro/m²] applied in this analysis.

Material	euro/m2
Zinc	21
Galvanised steel	25
Natural stone	235
Rendering, mortar	25
Paint	11

Effects of Air Pollution on Agricultural Systems

Geographically distributed data on crop production were obtained from ITE for the UK. Across the rest of Europe the data adopted were national totals for production of each crop type.

Acidification of agricultural soils

Lime is routinely applied to farmland to counteract acidification linked to farming practises, including harvest. Atmospheric deposition increases the amount of lime required to maintain acidity levels. The basis of the method applied here to calculate the costs associated with this change in demand for lime is as follows:

• The total amount of acidifying pollutant deposited to the land surface in a given area;
• The amount which falls on soils which require lime (excluding, for example, urban areas, water and soils on calcareous drifts);
• The cost of neutralising this amount of acidic deposition with lime;
• The increased acidic deposition in this area resulting from the change from one scenario to another;
• The additional cost of neutralising the difference in inputs to soils, which require lime (priced at 16.8 euro per tonne of lime).

Fertilisational effects of nitrogen deposition

Nitrogen is of course an essential plant nutrient, applied by farmers in large quantity to their crops. The deposition of additional nitrogen to agricultural soils is thus beneficial (assuming that the dosage of any fertiliser applied by a farmer is not excessive). The analysis quantifies total deposition of nitrogen to arable land and permanent pastures. The benefit is calculated directly from the cost of nitrate fertiliser, 430 euro /tonne of nitrogen (Nix, 1990). Given that additional inputs will still be needed under current conditions to meet crop N requirements for intensive agricultural systems there is a negligible saving in the time required for fertiliser application (if any), so it seems reasonable to cost benefits purely in terms of the (perhaps theoretical) reduction in N required as fertiliser. This analysis probably tends to overestimates the benefit of N deposition. N is deposited from the atmosphere throughout the year, including times when crops are not actively growing. The potential for deposited N to drain off and cause eutrophication is not monetised.

Similar analysis has not been performed for afforested areas. There is concern that prolonged deposition of N to these areas can lead to nutrient imbalance (Schulze et al, 1989), and hence that observed benefits in terms of enhanced productivity are not sustainable.

Ozone effects

Ozone damage to crops has been calculated using EMEP’s accumulated ozone above a threshold of 40 ppb (AOT40) metric, where AOT40 is defined by:

The time integral is over the growing season, which, for crops, is taken to be daylight hours in the months May-July. The metric has the units ppb.hours, or ppm.hours.

Functions are listed in tables AII.8 and AII.9.

Table A6.5. Estimated sensitivity of different crops to ozone. Species written in normal type are discussed in the review by Jones et al (1997). Species written in italics are not specifically discussed by Jones et al, but do feature in European crop production statistics. Sensitivity in these cases is estimated by analogy with similar crops.

Tolerant crops	maize raspberries cabbages	barley olives	leaf crops olive oil	sugar beet strawberries
Slightly sensitive crops	pasture grass rice	sorghum millet	oats	rye
Sensitive crops	wheat potato apples lemons limes flax hemp	clover tomato oranges peaches pears hops linseed	soybeans sunflower plums grapefruit tangerine onion rapeseed	beans grapes watermelons carrots cucumbers dates sesame seed
Very sensitive crops	tobacco

Table A6.6. Ozone exposure-response functions.

Crop type	Exposure Response Function
	% loss per ppm.hour AOT40
Tolerant crops	0
Slightly sensitive crops	1.0
Sensitive crops	1.75
Very sensitive crops	3.57
Meat and milk products	0.5

SO₂ effects

The following functions were used to quantify % yield change (y) from SO₂ effects on agriculture, derived from the work of Baker et al (1986), accounting for the fertilisational effect of sulphur at low concentration (European Commission, 1995);:

y = 0.74(SO₂) - 0.055(SO₂)² (from 0 to 13.6 ppb SO₂)

y = -0.69(SO₂) + 9.35 (above 13.6 ppb SO₂)

These functions have been applied to the following crops:

maize	barley	wheat	sorghum
oats	rye	millet	rice
leaf crops	sugar beet	raspberries	strawberries
soybeans	beans	potato	tomato
sunflower	carrots	cucumber	flax
hops	hemp	linseed	sesame seed
tobacco

For pasture the following function has been used, based on a review by Roberts (1984). All data used to derive the functions was taken from studies on Lolium perenne, the most common pasture grass in Europe. Again, the functions have been adapted to account for fertilisation of crops below the lowest exposure adopted experimentally.

y = 0.20(SO₂) - 0.013(SO₂)² (from 0 to 15.3 ppb)

y = -0.18(SO₂) + 2.75 (above 15.3 ppb)

Meat and milk production are assumed to be 50% as sensitive as pasture grass, on which livestock are primarily dependent for food.

Valuation of crop losses

Valuation of crop losses has been undertaken using prices from United Nations Food and Agriculture Organisation (FAO, 1994).

Ecosystem Damage

See Appendices V and VII.

Forest Damage

Two functions have been applied for assessment of ozone effects on forest productivity (Karenlampi and Skarby, 1996):

Species: beech; % productivity change = -0.27x

Species: Norway spruce; % productivity change = -0.18x

where x = ozone expressed as AOT40, (the ozone concentration accumulated over a threshold of 40 ppb in daylight hours over the growing season), expressed in ppm.hours.

An average is taken of the results of applying the two functions to all European forested areas. Given the uncertainties in derivation of the two functions it is not appropriate to provide separate quantification of impacts for hard and soft woods. The analysis probably overestimates the damages linked to forest damage from ozone, on the grounds that forest managers would probably respond to mitigate damage through alterations to management procedures over the lifetime of forests. However, there is also scope for underestimation of overall effects on forest productivity given that no economic assessment is made of the effects on forestry of acidification, N deposition and exposure to SO₂ due to a lack of data. Impacts are quantified only to the extent of identifying areas in which ecosystems (including forests) experience exceedence of critical loads and critical levels.

Appendix 5          Appendix 7