Morning vehicle-start effects on photochemical smog

The influence of cold-start vehicle emissions on air ... found to increase CO peak concentrations from ... tributions allocate emissions in directprop...
0 downloads 0 Views 864KB Size
CURRENT RESEARCH Morning Vehicle-Start Effects on Photochemical Smog Jose

R. Martinez,'

Richard A. Nordsieck, and Alan Q . Eschenroeder

General Research Corp., P.O. Box 3587, Santa Barbara, Calif. 931 05

The influence of cold-start vehicle emissions on air quality is investigated using a photochemical/diffusion model. Both the time and space distribution of cold starts are examined. A day from a n October 1968 Los .4ngeles smog episode serves 3s a baseline for determining meteorological and chemical parameters for the model. Emission inputs consist of vehicular and stationary sources for 1968. 1971%1974. and 1980. Stagnant conditions in the central Los Angeles basin are assumed. Cold-start emissions are tound to increase C O peak concentrations from 9-137~while increases in ozone and nitrogen dioxide are practically insignificant. Decentralizing the starts geographically produced no significant differences from the case of uniformly distributed starts. either in pollutant loading or in air quality. -

Time phasing of pollutant emissions might exert a powerful influence on air quality in the case of photochemical smog. During the morning hours this is possible because on most days a period of atmospheric stability and stagnation is followed by a rising trend of solar radiation input. klan!- of the analyses of motor vehicle air pollution contributions allocate emissions in direct proportion to carmile5 per hour or traffic intensities. However. suboptimum fuel mixture conditions during the vehicle start phase increase some of the emission rates above those for normal operating activities. The increases are larger for those starts occurring when the vehicle engine is not warm. Therefore. the morning starting contributions are especially significant since most of the vehicle population is subjected to start-up-about 9 0 7 ~of them between 6 and 9 X.hl. in Los Angeles (Kearin and Lamoureaux. 1969)--and most of them have been idle for several hours preceding the start. This hecomes a key, issue in the design of control system certification tests because the cold start is a relatively infrequent phenomenon and running emissions are continuously distributed. Because of the nature of the morning time phasing. it may be that a uniform proration of starting emissions over the whole day's activities is not a n adequate criterion for evaluating a control quently a n analysis of the time-phasing e sary to assess the cause effect relationship between startup contributions and air quality. Then the results of the anaiysii can be translated into a weighting scheme for 1

'To whom correspondence should be addre\sed

combining the results of cold-start cycle and hot -start cycle tests for evaluating the performance of proposed control schemes. In this report, the relative importance of morning coldstart emissions is assessed for those effects associated with photochemical smog. A computer simulation model iEschenroeder and Martinez. 1971. 1972) is used with ac'rometric data from the Los Angeles basin to study the buildup of air pollution a s it is affected by starting emissions. The procedure relates meteorological factors, time 'space traffic distributions. and ultraviolet solar radiation to the photochemical atmospheric mechanisms involved in air pollution.

t h od of A p p r o u c h Using validated input values and chemical parameters for a representative baseline case, we set out to determine air quality changes t h a t are ascribable to vehicle start-up. The first step in the investigation is a determination of the source emission inventory for stationary and mohile emitters. Peculiar to our purpose is the segregation of the temporal and geographical distribution of starting contributions from the running emission distributions. Then. selecting a stagnant day in the Los Angeles basin. we artificially remove these contributions to assess their effect on air quality. This is done for 1968. 1971. 1974. and 1980. Some of the years were chosen to make best use o f t h e e s isting data base and others to reflect substantial changes in emission control systems. After the source characterization. the rnodel is used to obtain a simulation of the baseline case. The day of October 23. 1968. was chosen as the base case because ir is typical of days with stagnant meteorological conditions and high cmog in the Los Xngeles Basin. The simulation involves mudeling the buildup of various pollutants. ( ' 0 . NO. S O * . 0 3 . and reactive hydrocarbon. at Huntington Park in the Los Xngeles Basin. The central basin location of Huntington Park makes it likely that automotive eniissions play a large role in producing the pollutant concentrations observed. Measurements made by Scott Research Laboratories (1969) serve as the data against which the computed concentrations are compared. T o follow a worst-case approach. no attempt is made to simulate advection a t Huntington Park. Additional details about the simulation of the base case are found in Martinez et al. (1971). The model itself has been described in Eschenroeder and Martinez (1971, 1972). In an attempt to capture realism, all our simulation results include appropriate stationary source emissions as well as mobile source emissions constructed from vehicle Volume 7 .

Number 1 0 , October 1973

917

age distributions. In keeping with the use of a stagnant slab model to approximate the central basin, we selected a baseline day with little wind, resulting in relatively high pollution. The possibility has been suggested that a nonuniform spatial distribution of vehicle starts may reduce the effects of the cold-start contribution. To test this hypothesis, a second case was simulated using a moving air parcel. An air trajectory was selected that would traverse the Los Angeles basin from west to east during the morning and early afternoon. The trajectory uses average September wind patterns and is typical for a September day which has low winds and stable atmospheric conditions. Using this trajectory, we compare the pollution levels for a single year under two different assumptions for the distribution of cold starts: a uniform and a nonuniform spatial distribution. The 1974 emissions data were used to generate source inputs for the trajectory. Successive sections describe the modeling of source emission distribution, the choice of a baseline case, and the simulation results. The final section contains some overall observations and cautions regarding the results. The chemical-kinetic model used has been described in Eschenroeder and Martinez (1971). Source Emission Distribution In general, contaminant source emissions are distributed in space and time in a rather complex manner. The model of geographical and hourly variations used here is based largely on the work of Roberts et al. (1971). The System Development Corp. (SDC) driving patterns survey (Kearin and Lamoureaux, 1969) provided a means for estimating the magnitude and temporal distribution of the morning start-up phenomenon, but the geographical distribution was chosen somewhat arbitrarily for lack of readily usable data. Annual variations in average daily source emissions, resulting from urban population growth and the increasing number of vehicles with various levels of emission control devices, were drawn from projections by the California Air Resources Board (ARB) (Air Resources Board, 1971) and Los Angeles Air Pollution Control District (LAAPCD) (Air Pollution Control District, 1969) and from available test data (Huls, 1971), standards, and surveys (McGraw and Duprey, 1971). Average Annual Emission Factors for Motor Vehicles. Table I shows recent test data from Huls (1971) for vehicle emission factors as measured on the new federal driving cycle (DHEW, 1970), with and without an initial cold start. The tabulated differences, multiplied by 7.45 miles (the distance that would be covered during one federal cycle), provide baseline values of average grams per cold start for the five model years shown. Lacking measured emission factors for other model years, we chose to use federal standards where available and “scale” California standards and prestandard estimates to the federal cycle to fill in any remaining gaps between 1960 and 1974. Pre-1960 emission factors were assumed to be constant a t the 1960 levels, and similarly, the 1975 emission factors were extrapolated as constant for all subsequent years. The scaling technique used is based on California ARB calculations of “equivalent standards” for 1972 under the new federal procedure (Air Resources Board, 1970). Table I1 is reproduced from that reference. Based on these data, the scale factors 3.2/1.5 and 47/23 were used (where necessary) to convert California standards and prestandard estimates for HC and CO emissions to equivalent federal emission factors. NO, emissions must be scaled slightly differently since the current 918

Environmental Science & Technology

Table I . Federal Cycle Gram-per-Mile Emissions Based on Measurements from Huls, 1971 NOx (asNO2)

Model 1968 ( 6 tests) 1969 .(54 tests) 1970 (57 tests) 1971 (25 tests) 1975= (3 tests)

6.11 5.35 0.76 5.28 4.40 0.88 6.18 5.39 0.79 4.43 3.71 0.72 0.85 0.99 -0.14

Cold Start Hot Start Difference Cold Start Hot Start Difference Cold Start Hot Start Difference Cold Start Hot Start Difference Cold Start Hot Start Difference

HC

co

4.80 3.36 1.44 4.79 3.51 1.28 4.11 2.65 1.46 3.63 2.80 0.83 0.61 0.25 0.36

71.34 37.23 34.11 48.14 28.87 19.27 41.91 22.67 19.24 42.08 29.66 12.42 86.68 0.90 5.78

a These three tests were carried out on 1971 model cars equipped with prototype versions of the emission control devices to be used on 1975 production models

Table I I. Equivalent California and Federal Emission Standardsa Pollutant HC

co a

California 1972 standards

Equivalent new procedure

Federal 1972 standards

1.5 g / m i 23 g / m i

3.2 g / m i 47 g/mi

3.4 g / m i 39 g/mi

Air Resources Board (Calif ) , 1970

federal test procedure does not provide for their measure,merit. However, we are able to determine federal-equiva-

lent NO, emissions using the NO, measurements from the California test procedures. California has augmented the test procedure by requiring that the federal test cycle be followed by two hot 7-mode (California) cycles for nitrogen oxides testing. The new equivalent 1972 standard under this procedure is 3.2 g/mi NO, as compared to 3.0 g/mi under the $-mode test procedure alone (Air Resources Board, 1970). Hence, the scale factor 3.213.0 was used to convert NO, emissions based on California cycle testing to their equivalent federal values. Once scaled, all emission factors were assumed to contain the effect of one cold start as specified in the federal procedure. To summarize, pre-1968 emission factors were obtained by the scaling technique, measured emissions data were used for the years 1968 through 1971 and for 1975, and the more stringent of the scaled California standards or the federal standards were ;sed for the years 1972, 1973, and 1974. Given data for the contributions of cold start t o the emission factor for the model years shown in Table I, we extrapolated back from 1968 and forward from 1975 assuming constant levels to obtain estimates of cold-start contributions to emission factors in other years. Between 1971 and 1975, the cold-start differentials were assumed to scale down in proportion to changes in exhaust emission standards in the intervening years. We then subtracted these cold-start contributions from the scaled federalcycle emission factors to yield estimates of “hot-running” emission factors for the years not included in the test data. Table I11 summarizes the “hot-running’’ emission factors and cold-start emissions used in the study. (As mentioned above, the gram-per-start cold-start emissions are

obtained by multiplying the gram-per-mile cold-start differentials by 7.45 miles.) The 2.69 g/mi evaporative emission factor shown in Table I11 for pre-1971 cars was calculated from data (Air Resources Board et al., 1969) by assuming an average gasoline Reid vapor pressure of 8.5 lb/ inz, a Los Angeles County auto population of 3.95 X 106 c a n (Air Pollution Control District, 1969), and an average annual mileage of 10,000 miles per car. The result is in good agreement with McGraw and Duprey (1971). The evaporative emission controls introduced on 1971 and later cars were assumed to perform as specified. Blow-by emissions on pre-1962 cars were set a t 4.1 g/mi (McGraw and Duprey, 1971) and were likewise assumed to be eliminated on post-1961 cars by the introduction of positive crankcase ventilation systems. The model year emission factors in Table I11 were then combined in accordance with a vehicle age and usage distribution (Hocker, 1971) shown in Table IV to yield weighted average vehicle emission factors for each year to be analyzed. The resulting average running and cold-start emissions for 1968, 1971, 1974, and 1980 are shown in Table V. Geographical Distribution of Motor Vehicles. Using an extensive data base of traffic counts in the Los Angeles area, Roberts et al. (1971), of Systems Applications, Inc. (SAI), have characterized the spatial distribution of Los Angeles freeway and surface street traffic intensities in each of 625 grid squares of 2 X 2 miles each. Average daily ~~~

traffic intensity within each square is represented by an estimate of the total number of vehicle miles (either freeway or nonfreeway) driven in that square. The absolute traffic intensities given by SA1 were assumed to apply as of 1968. Two geographical distributions of the total “pulse” of cold-start emissions were used. For the stagnant-slab model, we assumed that morning start-ups were distributed uniformly, using the area of the Los Angeles Basin (1250 milesz-Air Pollution Control District, 1969) as a n estimate of the size of the densely populated region. One of the primary purposes of analyzing a cross-basin trajectory was to evaluate the effect of a more suburb-oriented cold-start distribution. Hence, for that case, we modeled both the uniform cold-start distribution and one in which the density of morning start-ups (assumed equivalent to car residence density) was assumed to be three times as high at the outer edge of the populated area as in the basin center (here taken as the Federal building downtown), varying linearly in between. Temporal Distribution of Traffic. The hourly distributions of average daily traffic intensities for freeway and nonfreeway traffic used in the study were those developed by SA1 in conjunction with the geographical distributions described above. In their report (Roberts et al., 1971), SA1 shows that only small errors are incurred by characterizing all freeway traffic by one time distribution and all nonfreeway traffic by another.

~

Table 1 1 1 . Hot-Running and Cold-Start Emissions by Model Year Cold-start emissions, g/start

Hot-running emissions, g/mi Model year

NOx (as NOz)

HC

co

NOx (as NO?)

HC

1960 and

3.50a

22.06O 2.69 evap. 4.1 0 blow-by 22.06a 2.69 evap. 4.10 blow-by 22.06O 2.69 evap. 22.06O 2.69 evap. 22.06a 2.69 evap. 22.06O 2.69 evap. 5.81a 2.69 evap. 5.81a 2.69 evap. 3.36b 2.69 evap. 3.51 2.69 evap. 2.65h 2.69 evap. 2.80b 2.63a 2.63a 2.63a

1 28.9O

5 66

10.73

254.1

1 28.ga

5.66

10.73

254.

128.ga

5.66

10.73

254.

128.ga

5.66

10.73

254.

128.ga

5.66

10.73

254.

128.9O

5.66

10.73

254.

35.39a

5.66

10.73

254.

35.39a

5.66

10.73

254.

37.23b

5.66h

10.73h

254.

28.87h

6.56b

9.540

143.6b

22.67b

5.8gb

10.88b

143.3h

29.66b 28.7OC 28.7OC 28.7OC

5.36h 4.02& 3.78d 3.78d

6.18b 4.25d 4.25d 4.25d

92.53b 76.74d 76.74d 76.74d

- 1.04e

2.68b

43.06h

before 1961

3.50a

1962

3.50a

1963

3.50a

1964

3.50a

1965

3.50a

1966

5.64a

1967

5.64a

1968

5.350

1969

4.40h

1970

5.390

1971 1972 1973 1974 1975 and

3.71h 2.66a 2.4gC 2.4gC

after

0.990

O.2Sb

0.90b

co

I

Obtained by scaling California standard or estimate of prestandard emission factor to federal test procedure and subtracting cold-start contribution as extrapolated from test data (see text) Test data (see Table I ) . Obtained by subtracting cold-start contribution from federal standard. Lacking data in the 1972-74 time period. i t was assumed that the techniques ernployedto meet the standards used in obtaining the running emission factors (see a or c above) would act proportionally in scaling the cold-start emissions down from their 1971 values toward their 1975 values. e Test data. The negative correction is always used in conjunction with the running emission factor, thereby reducing each post-1975 car’s net NOx contribution to the level Dredicted for a federal cvcle with one cold start, followed bv hot-runnina emissions.

Volume 7 , Number 10, October 1973 919

Data reported (Kearin and Lamoureaux, 1969) on the distribution of weekday trip start times in Los Angeles indicate that the temporal distribution of morning car starts can reasonably be approximated by a triangular pulse. The model selected starts from zero a t 6:OO A.M., rises linearly to a peak a t 7:30 A . M . , and then falls linearly back to zero at 9:00 A.M. The data also showed that of an average 4.4 trips per car weekday (in Los Angeles), 20.6% or 0.907 trips per car weekday were started between 6:OO and 9:00 A . M . Thus about 91% of the registered automobiles in Los Angeles County were assumed to contribute one morning start to the cold-start contaminant pulse.

'ARK

\,

i

Figure l . Cross-basin air trajectory

Table I V . Vehicle Age and Usage Distribution at Close of Model Yeara (From Hocker, 1971) Age. years

Fraction of population

Miles driven in last year

0.108 0.105 0.102 0.098 0.093 0.088 0.081 0.072 0.062 0.051 0.140

15,000 13,000 11,000 9,600 8,400 7,000 5,300 5,000 4,400 4,200 3,500

1 2 3 4 5 6

7 8 9 10 over 10

a Based on a survey of California vehicles stopped at California Highway Patrol roadside inspection stations.

Table V. Average Vehicle Emission Factors Hot-running emissions sim1 Year

N Ox (as NO>)

1968 1971 1974 1980

4.51 4.55 3.55 1.52

920

Cold-start emissions. g, start

HC

co

N Ox (asNO2)

HC

co

16.81

82.49 51.58 37.80 6.55

5.66 5.76 4.81 C.17

10.73 9.69 6.89 3.46

254.10 188.71 128.05 54.69

10.00 5.70 0.93

Environmental Science & Technology

Annual variations of total pollutant emissions from motor vehicles were assumed to grow in proportion to the registered vehicle population. Hence, based on an extrapolation of vehicle registration in Los Angeles County (Air Pollution Control District, 1969) and using 1968 as the baseline case, the vehicle mileages in each square of the geographical distributions were increased accordingly for each year analyzed beyond 1968. Of course, the cold-start contaminant pulse was similarly scaled up. Stationary Source Emissions. The stationary source emissions modeled for this study included oxides of nitrogen and reactive hydrocarbons as characterized by Roberts et al. (1971). Stationary carbon monoxide emissions were neglected in comparison with motor vehicle CO emissions. Their model uses the same 25 X 25 grid of 2 x 2-mile squares described above to provide geographic source distribution. Within each square, the flux is given in kilograms per hour and is assumed to be constant between 6:OO A.M. and 6:OO P.M. These data were assumed to be circa 1968, and subsequent annual growth was modeled on the basis of ARB projections for the South Coast Air Basin (Air Resources Board, 1971). The ARB projections show negligible expected changes in total stationary hydrocarbon fluxes, but increases in stationary NO, fluxes totaling 37% by 1980. Appropriate factors were used to increase the kilogram-per-hour NO, fluxes in each grid square for each year modeled. Input for Central Basin Stagnant Case. For the central basin stagnant-mixing case, it would seem inappropriate to simply use the source data specified for the single 4-square-mile grid square containing the central-basin measurement station (Scott Research Laboratories, 1969). Instead, to simulate more extensive local mixing during the 8-hr period of interest, source emission fluxes in neighboring grid squares were combined in a n areaweighted average to yield a n input flux history which represents an average over a 5 x %mile square centered on the measurement station. Input for Cross-Basin Trajectory Case. The Lagrangian air parcel formulation used in the current General Research Corp. model has been described elsewhere (Eschenroeder and Martinez, 1972). For the cross-basin trajectory case. the data necessary to determine the schedule of changes in pollutant fluxes into the air parcel on a realistic cross-basin trajectory were obtained from Neiburger and Edinger (19543. Figure 1 shows the trajectory obtained from a combination of average September wind speeds and directions measured a t stations near the coast, where the trajectory starts, and downtown a t the Federal building, where the trajectory passes a t about noon. The afternoon portion of the trajectory was obtained by estimating wind speeds and directions from a similar trajectory in Neiburger and Edinger (1954). The resulting trajectory is very similar in path and time history to one constructed by Neiburger and Edinger from more detailed analysis of average hourly resultant wind streamlines for September, and is therefore considered to he a realistic cross-basin air trajectory. Because the wind speeds near the coast are practically zero early in the morning, the trajectory shows no appreciable movement until after 8:OO A.M. As the air parcel subsequently moves along the trajectory, the schedule of pollutant influxes is determined by both the geographical and temporal source distributions described above.

Establishment of Baseline Case

a

T o provide data base against which to compare the results of the atmospheric model, we selected a day dur-

ing a smog episode, October 23, 1968. For this date, we used data recorded by Scott Research Laboratories (1969) a t Huntington Park. Because Huntington Park is centrally located in the Los Angeles Basin, its aerometric data are likely to reflect the effects of vehicular sources. This is an important consideration in this study. The day of October 23 was chosen for several reasons: Aerometric and meteorological data are readily available. The day is typical of high-oxidant, heavy-smog days in Los Angeles. The prevalence of low-wind speeds up to about 9:30 A.M. and the presence of a low inversion base (220 meters) indicate stable atmospheric conditions which lead 'to a worst-case approach in our computed results. The authors have previous modeling experience with this date (Eschenroeder and Martinez, 1969a). The baseline case was used to determine the parameters of the model related to meteorological conditions, chemical reactivity, and nitrogen balance (Eschenroeder and Martinez, 1970; Gordon et al., 1968; Gay and Bufalini, 1971). Full details of the simulation of the baseline case are found in Martinez et al. (1971). The intent is to model pollutant concentration buildups a t a single location on a specific day. The simulation of the baseline case thus provides a reference set of model parameters which reflect the prevailing meteorological conditions and the reactivity of the air a t that location. Once set, these parameters remain frozen in subsequent exercises of the model which differ only in the strength of the source emissions. Thus the results presented in the next two sections were obtained using the same basic framework in all cases; any differences reflect changes in emissions only. Finally, following a worst-case approach, a stagnant slab model was used. Four years were chosen for the cold-start vs. hot-start comparison: 1968, 1971, 1974, 1980. The first year, 1968, has extensive atmospheric data used for validation as mentioned above, but the accuracy of the vehicular emission data base can only be classified as fair due to the large quantity of pre-1966 cars still on the road in 1968. By contrast, 1971 has the best and most reliable auto emissions data base of all the years and it was chosen for this reason. The year 1974 was selected because it is the last model year before automotive emissions must satisfy the stringent 1975 standards. Thus for 1974, effects of the cold-start pulse are unlikely to be overshadowed by emissions from stationary sources. Finally, by 1980 over 70% of all car-miles in Los Angeles should be traveled by cars with 1975 or post-1975 emission controls. Hence 1980 was chosen because it is the year when the full impact of the 1975 auto emission standards is supposed to be felt.

ground concentration with cold start divided by ground concentration without cold start, taken a t 1300 hr (the end of the simulation run) for 0 3 and NOa, and a t peak concentration (whenever it occurs) for CO. Table VI11 combines carbon monoxide data from Tables VI and VI1 with the addition of the CO emissions ratio for the 0600-1300 time period. Two features of these data are notable: first, cold-start contributions to total CO emissions are considerable, even when taken over the complete 7-hr span of the simulation; and second, in spite of a sizable increase in CO emissions in the relatively short vehicle start-up period, only modest increases in peak CO concentration are seen, giving us a measure of the mitigating. effects of atmospheric dispersion and dilution on nonreacting species. C o l d - S t a r t Effects: C r o s s - B a s i n T r a j e c t o r y C a s e

To determine the effect of a suburb-oriented cold-start distribution, the moving-air-parcel model was employed to compute various pollutant concentrations along a crossbasin trajectory typical of September wind patterns. When 1974 emissions data were used, two cold-start geographical distributions were simulated, one uniform and one weighted 3 to 1 between the suburbs and downtown, as discussed earlier. The cold-start contributions to pollutant loading in the air parcel are shown in Table IX as the ratios of emissions with cold start t o emissions without cold start for NO, CO, and reactive hydrocarbons. Note that the decentralized start distribution loads the air parcel with more pollutants than the uniform one because of relatively high morning exposure of the air parcel to areas away from the basin center.

Table V I . Emission Effects (Ratios of emissions with cold start to emissions without cold start, each for 0600-0900 time interval in central L.A. basin) Year Species

1968

1971

1974

1980

co

1.153 1.479

1.157 1.568

1.156 1.526

1.009 2.297

Reactive hydrocarbon

1.077

1.103

1.105

1,107

NO

Table V I I . Air Quality Effects (Ratios of concentration with cold start to concentration without cold start in central L.A. basin)

C o l d - S t a r t E f f e c t s :S t a g n a n t - B a s i n Case

The contribution of cold-start emissions to air pollution was determined for a location in the central Los Angeles basin on a day characterized by very light winds. concentration histories of various pollutants were computed, both with and without the morning vehicle start-up emissions in the 0600-0900 hour time interval. The effect of cold-start emissions of S O , CO, and reactive hydrocarbon on atmospheric loading during the early morning hours is demonstrated in Table VI in the form of emissions ratios for the 25-square-mile test region; i.e., ratios of emissions with cold-start to emissions without cold start. To show the effect of cold-start emissions on ambient air quality, we have computed concentration ratios for two product pollutants, ozone and nitrogen dioxide, and for carbon monoxide. Shown in Table VII, the ratios are

Year Time

Species

1 3 0 0 hours 1300hours Peak

O3 NO2 CO

1968

1971

1974

1980

1.024 1.011 1.093

1.028 1.015 1.132

1.014 1.014 1.127

1,071 1.002 (nopeak)

Table V I I I . Carbon Monoxide Effects (Ratios of quantities with cold start to quantities without cold start) Year

0600-0900 emissions 0600-1300 emissions Peak concentration

1968 1.479 1.21 6 1.093

1971

1974

1980

1.568 1.257 1.132

1.526 1.238 1.127

2.297 1.586

(no peak)

V o l u m e 7, N u m b e r 10, October 1973

921

Table IX. Emission Effects for 1 9 7 4 Trajectory (Ratios of emissions with cold start to emissions without cold start, each for 0600-0900 time interval) Spatially uniform start distribution

Decentralized start distribution

co

1.168 1.546

1.183 1.594

Reactive hydrocarbon

1.134

1.146

Species

NO

Table X. Air Quality Effects for 1 9 7 4 Trajectory (Ratios of concentration with cold start to concentration without cold start)

Time

Species

1400 hours 1400 hours

03

Peak

co

NOz

Spatially uniform start distribution

Decentralized start distribution

1.039 1.024 1.125

1.042 1.026 1.136

Table X I . Carbon Monoxide Effects for 1 9 7 4 Trajectory (Ratios of quantities with cold start to quantities without cold start)

emissions emissions Peak concentration

0600-0900 0600-1400

Spatially uniform start distribution

Decentralized start distribution

1.546 1.250 1.125

1.594 1.273 1.136

The effect of vehicle start distribution on air quality in the air parcel may be seen in Table X by comparing the concentration ratios computed for the uniform and decentralized start distributions. These concentration ratios (ground concentration with cold start to ground concentration without cold start) were computed a t peak value for CO, and at the end of the air-trajectory (1400 hr) for the chemical product pollutants 0 3 and NO2. Finally, as in the previous section, we compare the contributions of cold-start CO emissions t o total CO emissions, and to the resultant peak concentrations of CO in the air parcel (see Table XI). (The 0600-1400 time period is the complete duration of the simulated trajectory.) As before, the noteworthy features of these data are the considerable size of the cold-start emission contributions and the weakness of the coupling between emissions and air quality effects, even for a nonreacting species. The results of this investigation can be used to determine a weighting scheme for combining cold-start and hot-start driving-cycle measurements. (The use of a weighting factor to determine emission standards is defined in EPA Regulations, 1971.) It is apparent, however, that the time phasing of pollutant emissions must be considered in assessing the effects on air quality of the various phases of the driving cycle; simply averaging coldstart emissions throughout the whole day does not give a true picture of the cause/effect relationship between coldstart emissions and air quality.

Conclusions Considering the wide variety of emission sources and control measures, we find that the pollution input of cold starts can be very large or very small, as shown in Table VI. The nitric oxide contribution (between 6 and 9 A.M.) 922

Environmental Science & Technology

stays near 15% of all emissions until the end of the present decade, when it drops below 1%.Dominant effects of stationary sources compared with vehicles cause this sharp decrease. Reactive hydrocarbon starting emissions comprise about 10% of the 6-9 A.M. inputs. Because of choked engine operation, carbon monoxide-starting contributions range from 48-13070 of running CO emissions during the morning hours of peak traffic. It is of central interest here to examine the incremental change in air pollution due to the vehicle starting process. As illustrated in Table VII, 1300-hr levels of oxidant species, 0 3 and NOz, are practically unaffected within the uncertainties of the calculation. Apparently, the reaction times to produce these compounds are long enough to allow considerable atmospheric dilution. Carbon monoxide is treated as nonreactive in the simulations and it is more strongly affected than the oxidant species. Because carbon monoxide is emitted directly and oxidant is not, it may be more meaningful to use the emission ratios (Table VI) instead of the air quality ratios (Table VII) for setting standards. This procedure would more nearly reflect the increased exposure of receptors in the area where the cars are started. Expressed as percentages, cold-start emissions (Table VI) exhibit a long-term upward trend. This tendency is attributable to lower running emissions achieved by more sophisticated control systems, accompanied by thermal inertia in catalytic devices, which aggravates the coldstart problem in advanced systems. The oxides of nitrogen actually assume a lesser role as time goes on because they pose less of a cold-start control problem than do the other contaminants. Compounds of KOx are also special in that stationary inputs are projected to grow significantly over a period when vehicles are experiencing more stringent controls. Tables VI through VI11 demonstrate that the time phasing of vehicle activities and reactions influences the morning startup effects. The relatively high deposition rates in the 6-9 A.M. time interval have these effects superimposed on them. Spreading the averaging period over the 6 A.M. to noon time interval reduces the relative fractional contribution. Although it was not modeled, delaying the startup time and the traffic peak would also drive the air pollution levels downward, because higher wind speed and surface heating would enhance the dispersion. Spatial distribution of cold starts was also investigated in the study. An air trajectory, going from the western Los Angeles basin over the downtown area and up into Burbank, was chosen. Essentially no difference was noted between one case with uniformly distributed cold starts and another case with three times the cold-start density a t the coast as that in downtown (Tables IX-XI). This lack of sensitivity to spatial distribution occurred for all pollutants for the 0600-1400 time interval covered by the trajectory. Some results presented in this work run counter to intuition. This is common when one works with complicated nonlinear models. However, in view of all the uncertainties inherent in using a model to simulate smog, a question might be raised regarding the confidence that we can place on the results presented in this work. To a certain extent, results which show small perturbations of the base case give us some reassurance that the model is not being exercised beyond the limits established by the validation data base. On the other hand, the presence of large changes in the predictions warns us of the possibility that the validation limits are being exceeded. Confidence in the predicted cold-start effects is thus reinforced by the

smallness of these effects. Nevertheless, the results shown should be interpreted in terms of indicative trends rather than in terms of specific numbers.

Literature Cited Air Pollution Control District (County of Los Angeles), “Profile of Air Pollution Control in Los Angeles County,” 1969. Air Resources Board (Calif.), “Air Pollution Control in California: 1970 Annual Report,” 1971. Air Resources Board (Calif.), Los Angeles County Air Pollution Control District, and Western Oil and Gas Association, “Gasoline Modification: Its Potential as Air Pollution Control Measure in Los Angeles County,” p 28, 1969. Air Resources Board (Calif.), “Test Procedures, Emission Standards Revised,” Calif. Air Res. B d . Bull., Vol. 2, No. 9, p 4, Nov. Dec., 1970. “DHEW Urban Dynamometer Driving Schedule,” Fed. Regist., 35 (219), Part I1 App. A, 17311 (1970). “EPA Regulations,” ibid., 36,22456 (1971). Eschenroeder, A. Q , , Martinez, J . R., “A Modeling Study to Characterize Photochemical Atmospheric Reactions in the Los Angeles Basin Area,” General Research Corp. CR-1-152, 1969. Eschenroeder, A. Q., Martinez, J . R., A d u a n . C h e m . S e r . , 113, 101-68 (1972). Eschenroeder, A. Q., Martinez, J. R., “Analysis of Los Angeles Atmospheric Reaction Data from 1968 and 1969,” General Research Corp. CR-1-170, 1970. Eschenroeder, A. Q., Martinez, J . R., “Further Development of the Photochemical Smog Model for the Los Angeles Basin,” General Research Corp. CR-1-191, 1971. Available from Nat. Tech. Inform. Serv. Springfield, Va., Document No. P B 201737.

Gay, B. W., Bufalini, J. J., Enuiron. Sci. Technol., 5,422 (1971). Gordon, R., Mayrsohn, H., Ingels, R., ibicl., 2, 1117 (1968). Hocker, A. J., California Air Resources Board, 9528 Telstar Ave., El Monte, Calif. 91731, private communication, 1971. Huls, T., Environmental Protection Agency, Air Pollution Control Office, 2929 Plymouth Road, Ann Arbor, Mich. 48105, private communication, 1971. Kearin, D. H., Lamoureaux, R. L., “A Survey of Average Driving Patterns in the Los Angeles Urban Area,” System Development Corp. TM-(L)-4119/000/01 (1969). Martinez, J . R., Nordsieck, R. A,, Eschenroeder, A. Q., “Morning Vehicle-Start Effects on Photochemical Smog,” General Research Corp. CR-2-191, June 1971. Available from Nat. Tech. Inform. Serv., Springfield, Va., Document No. PB203872. McGraw, M. J., Duprey, R. L., “Air Pollutant Emission Factors” (preliminary document), p 24, U.S.Environmental Protection Agency, 1971. Neiburger, M., Edinger, J. G., “Meteorology of the Los Angeles Basin with Particular Respect to the ‘Smog’ Problem,” Air Pollution Foundation Report So. l , 1954. Roberts, P . J. W., Roth, P. M., Nelson, C. L., “Development of a Simulation Model for Estimating Ground Level Concentrations of Photochemical Pollutants,” Appendix A, Systems Applications, Inc., Report 71-SAI-6, 1971. Scott Research Laboratories, “Final Report on Phase I, Atmospheric Reaction Studies in the Los Angeles Basin,” Vols. I and 11, 1969.

Receiced for reuieu M a r c h 7, 1972. Accepted June 15. 1973 Work supported by t h e Air Pollution Control Office of t h e Encironrnental Protection Agency under Contract N o . EHSD-71-22.

Kinetics of Sulfation of Reduced Alunite N. Gopala

Krishnan and Robert W. Bartlett’

Process Metallurgy Group, Department of Applied Earth Sciences, Stanford University, Stanford, Calif. 94305.

The United States contains more than 500 million tons of the mineral alunite, K~S04.3Alz(SO4)3.6HzO,which is a potential regenerable absorbent for removing SO2 from dilute effluent gas streams. The alunite must be desulfurized to Kz0.3A1203 prior to sorption in a process step identical with absorbent regeneration. The sulfation kinetics of hydrogen desulfurized alunite were determined between 100” and 500°C in SO2 concentrations from 0.10.5%. Sulfation below 300°C is limited to the reaction with KzO. Above this temperature, A1203 is also sulfated, resulting in the formation of potassium aluminum sulfate, and the rates can be expressed as parallel processes. Each sulfation rate is proportional to the SO2 concentration and the remaining unreacted oxide, KzO and A1&, respectively. The experimental activation energy is low. The sulfation of K 2 0 is independent of oxygen pressure and the rate-controlling step appears to be formation of potassium sulfite which is subsequently oxidized to sulfate. The sulfation rate of A1203 is proportional top(02)’Iz. H

Fuel combustion is the chief source of air pollution by sulfur dioxide, and about 18 million tons of sulfur dioxide are being emitted annually by power generating plants in the Cnited States (Bienstock and Field, 1960). Both liquid scrubbing techniques and sorption by dry solids are

* To whom correspondence should be addressed.

being considered for the removal of sulfur dioxide that occurs in dilute concentration in power plant gases. Dry, elevated temperature sorption processes retain the buoyancy of the stack gases for maximum dispersion in the atmosphere after treatment. Among the dry processes are absorption by metal oxides to form sulfites or sulfates, with subsequent regeneration of the oxide and recovery of sulfur in a concentrated form. Alkalized alumina (NazO.Al203) developed by the U.S. Bureau of Mines is an excellent regenerable sorbent for sulfur dioxide (Bienstock et al., 1967). The mineral alunite, a hydrated potassium aluminum sulfate (Kz0.3A1203-4S03.6HzO) is chemically similar to the final sulfated product of alkalized alumina. Heating alunite a t 700°C decomposes the aluminum sulfate to oxide (Fink et al., 1931). Hydrogen reduction converts all the sulfate to hydrogen sulfide and oxides. The oxides thus generated react with sulfur dioxide and oxygen t o form sulfates. Alunite is found in several locations in the western United States and the estimated reserves recoverable by inexpensive surface mining are over 500 million tons ( E n g Mining J , 1972). The purpose of the present study was to investigate the sulfation kinetics of hydrogen desulfurized alunite.

Materials and Experimental Method Pelletized alunite from Marysvale, Utah, was used in this study. The mineral was crushed t o -100 mesh, mixed Volume 7, N u m b e r 10, October 1973

923