EPA-450/3-75-068
July 1975
IMPACT
OF ENERGY SHORTAGE
ON AMBIENT SULFUR DIOXIDE
AND PARTICULATE LEVELS
IN METROPOLITAN BOSTON
AQCR
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Wasle Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
EPA-450/3-7 5-068
IMPACT
OF ENERGY SHORTAGE
ON AMBIENT SULFUR DIOXIDE
AND PARTICULATE LEVELS
IN METROPOLITAN BOSTON
AQCR
by
R. Siegel, P. Guldberg, K. Wiltsee, andR. D'Agostino
Walden Research Division of Abcor, Inc.
201 Vassar Street
Cambridge, Massachusetts 02139
Contract No. 68-02-1830
Program Element No. 2AC129
EPA Project Officer: Gerald L. Gipson
Prepared for
ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
July 1975
-------
This report is issued by the Environmental Protection Agency to report
technical data of interest to a limited number of readers. Copies are
available free of charge to Federal employees, current contractors
and grantees, and nonprofit organizations - as supplies permit - from
the Air Pollution Technical Information Center, Environmental Protection
Agency, Research Triangle Park, North Carolina 27711; or, for a
fee, from the National Technical Information Service, 5285 Port Royal
Road, Springfield, Virginia 22161.
This report was furnished to th^ Environmental Protection Agency
by Walden Research Division ot Abcor, Inc. , Cambridge, Massachu-
setts 02139, in fulfillment of Contract No. 68-02-1830. The contents
of this report are reproduced herein as received from Walden Re-
search Division of Abcor, Inc. The opinions, findings, and conclusions
expressed are those of the author and not necessarily those of the
Environmental Protection Agency . Mention of company or product
names is not to be considered as an endorsement by the Environmental
Protection Agency.
Publication No. EPA-450/3-75-068
11
-------
TABLE OF CONTENTS
Secti on Title
I INTRODUCTION 1-1
A. PROBLEM BACKGROUND .' 1-1
B. PROGRAM OBJECTIVE 1-2
C. SUMMARY OF RESULTS AND CONCLUSIONS 1-3
II AIR QUALITY ANALYSIS 2-1
A. BACKGROUND 2-1
B. DATA ANALYSIS 2-18
C. STATISTICAL INFERENCE 2-22
III REGULATORY AND EMISSIONS ANALYSIS 3-1
A. OBJECTIVES AND SCOPE 3-1
B. METHODOLOGY 3-2
C. RESULTS 3-6
IV MODELING ANALYSIS 4-1
A. OBJECTIVES AND SCOPE 4-1
B. METHODOLOGY 4-1
C. RESULTS 4-6
V REFERENCES 5-1
VI APPENDICES
APPENDIX A - AIR QUALITY ANALYSIS A-l
APPENDIX B - EMISSIONS AND REGULATORY ANALYSIS . B-l
APPENDIX C - MODELING ANALYSIS C-l
lUhUeni
-------
ACKNOWLEDGEMENTS
This work was supported under Contract No. 68-02-1830 by the Environ-
mental Protection Agency (EPA). The assistance and guidance of the EPA
Project Officer, Mr. Gerald L. Gipson, and the assistance of Mr. Kenneth
Hagg and Mr. Thomas Parks of the Bureau of Air Quality Control of the
Commonwealth of Massachusetts were greatly appreciated.
IllUen,
-------
I. INTRODUCTION
A. PROBLEM BACKGROUND
Demand for energy in the United States has increased at an average
annual rate of 4.3 percent in recent years while per capita demand has
grown at a corresponding rate of 3.5 percent [1]. Both rates are signi-
ficantly higher than long-term historical averages. The bulk pf this
increased demand for energy has been met by increased use of liquid fos-
sil fuels. Consumption of oil increased each year since 1970 at an av-
erage annual rate of more than a million barrels per day (mbd) to over
17 mbd by end of 1973.* During this same period, oil production by dom-
estic sources and traditional foreign suppliers (Canada and Venezuela)
decreased, forcing an increased reliance on oil imports from the Middle
East. The American Petroleum Institute reports that foreign oil imports
accounted for 36.9 percent of the oil consumed in the United States in
1974. This increased dependency on foreign suppliers placed the United
States in an extremely vulnerable position in October 1973 when the
politically motivated Middle East oil embargo began. The embargo dras-
tically restricted oil imports into the country throughout the winter of
1973-1974 leading to severe supply problems in many sections of the country
during the critical winter heating period. In addition, the embargo
caused the price of imported oil to more than triple in a year's time.
This situation has come to be known as the "energy crisis" or the "energy
shortage" of 1973-1974.
Prior to the oil embargo, as part of State Implementation Plans
to attain national ambient air quality standards for sulfur dioxide (S02)
and total suspended particulates (TSP), several states adopted regulations
limiting the sulfur and ash content of fuels burned in their jurisdictions.
During the winter of 1973-1974, the Environmental Protection Agency (EPA)
granted several temporary variances on these regulations due to restricted
*
Total petroleum demand fell 3.3 percent in 1974 from 1973 levels.
1-1
Maiden
-------
supplies of low sulfur oils. In some cases, fuel utilization facilities
were permitted to convert from oil to coal usage to avoid complete cur-
tailment of their activities. Such changes were expected to cause in-
creased emissions of S02 and particulates, while, by contrast, concomi-
tant conservation efforts were expected to decrease fuel consumption and,
therefore, emission of these pollutants. Thus, the impact on ambient S02
and TSP levels, of the fuel changes necessitated by the energy shortage,
is not readily apparent. Observed variations in meteorological condi-
tions between the winter of 1972-1973 (no shortage) and the winter of
1973-1974 (shortage) further complicates resolution of the question of
the impact on ambient air quality of the energy shortage.
B. PROGRAM OBJECTIVE
The purpose of this project was to evaluate the impact of the
1973-1974 energy shortage on ambient air quality on a case study basis
for a major urban area, Metropolitan Boston. Three principal tasks were
undertaken to achieve this objective: an air quality analysis, a regu-
latory and emissions analysis, and a diffusion modeling analysis.
The objective of the air quality analysis was to identify,
quantify and interpret trends or changes in trends in measured S02 and
TSP levels during the period January 1970 through March 1974. The prin-
cipal focus of the analysis was to statistically test the relation be-
tween trends in regional air quality levels,and SIP regulations and vari-
ances granted because of the energy shortage. Other factors simultaneously
influencing the measurements, such as meteorological conditions, were
also evaluated.
• The objective of the regulatory and emissions analysis was to
develop quarterly emissions and fuel use inventories for S02 and particu-
lates in the Metropolitan Boston Air Quality Control Region (AQCR) for
the period January 1973 to June 1974 to support the subsequent modeling
analysis task. These data were developed to reflect gross changes in fuel
use between the base case (1972) and the periods of interest related to
growth, conservation measures, meteorological factors such as degree-days,
and variances granted and implemented on the Massachusetts SIP regulations.
UJalden
-------
The objective of the diffusion modeling task was to obtain a
clearer understanding of the impact and potential impact of the energy
shortage on air quality levels throughout the AQCR by isolating the rela-
tive effects of factors such as meteorology, growth and conservation, and
variances on ambient S02 and TSP concentrations. The results were used
to corroborate the findings of the analogous statistical decomposition of
measured air quality data undertaken in the air quality analysis.
C. SUMMARY OF RESULTS AND CONCLUSIONS
A combination of statistical analysis of ambient monitoring data
and simulation analysis based on diffusion modeling was used to examine
the effects of the energy shortage on sulfur dioxide (S02) and total sus-
pended particulate (TSP) levels throughout the Metropolitan Boston AQCR.
Changes in the emission of these pollutants associated with the energy
shortage during the first quarter of 1974 arose principally from fuel
conservation efforts, variances from existing sulfur in fuel regulations,
and conversion of the Salem Harbor power plant to coal-firing from oil-
firing. The findings of the study relative to these and other factors are
as follows.
1. The modeling analysis indicated that the energy shortage was
responsible for average concentration levels lower than those projected
to occur if historical fuel use trends had continued during the winter of
1973-1974. This result is a consequence of conservation measures' being
more pronounced than the opposing effect from implementation of variances.
Modeling analysis indicates that, in the absence of the energy shortage,
S02 concentrations would have risen an average of approximately 3 percent
across the AQCR between the first quarter of 1973 and the first quarter
of 1974, due to the combined effect of growth and changes in meteorology.
The analysis further indicates that the decrease observed across the AQCR
in S02 concentrations was approximately 12 percent, reflecting a net de-
crease of 15 percent due to the energy-shortage-related parameters. Fuel
conservation contributes approximately 18 percent to the decrease, while
variances contribute to an increase of 3 percent.
Uhlden,
-------
2. A major factor in the balance of energy shortage effects is
that the actual implementation of variances by combustion facilities in
the AQCR represents realization of only a small fraction of the potential
for consumption of higher sulfur fuel oil. This result was determined from
a survey conducted during the study of fuel distributors in the Region.
Simulation modeling indicated that if the variances had been fully imple-
mented, S02 levels would have deteriorated significantly across the AQCR
in spite of conservation effects.* The analysis indicated a potential
52 percent average increase in average S02 concentrations across the AQCR
associated with full implementation of the variances. The combined effect
of full variance implementation, meteorology, and growth and conservation
would have led to a net increase of 37 percent across the AQCR. Further,
this potential increase would have been highly variable across the region,
ranging from a net decrease of 6 percent southwest of Boston to a net in-
crease of 57 percent along coastal exposures. The impact of full variance
implementation is primarily due to a projected larger source contribution
from power plants and distillate area source users. Their combined contri-
bution would have been more than 90 percent of the total increase in SOo
levels due to variances.
3. The modeling analysis indicates that TSP concentrations in the
AQCR were largely unaffected by the energy shortage. Full implementation
of variances would not have significantly affected this finding.
4. The statistical analysis of air quality monitoring data be-
tween the first quarter of 1973 and 1974 also indicates that the influence
of meteorology on TSP and SC^ levels involved two opposing effects: the
tendency for the change in wind-stability patterns between these years to
decrease concentrations, and the effect of degree-days to increase concen-
trations. SC^ concentrations in the AQCR which are strongly related to
space heating requirements (degree-days) showed a net balance between these
factors. TSP concentrations, however, are affected considerably less by
* Assuming no additional conservation efforts concommitant with full variance
implementation.
/
Ulalden
-------
changes in degree-days and without this "balancing" factor showed a net
decrease due to changes in wind-stability patterns. Urban core TSP con-
centrations are more strongly influenced by space heating requirements,
due to the predominance of dwelling units and commercial establishments,
and thus showed a stronger balance between these factors.
5. The statistical analysis of air quality data also revealed
that S02 concentrations in the AQCR did not significantly increase during
the energy shortage period in spite of the partial implementation of the
variances granted on the State Implementation Plan to burn higher sulfur
fuel in several cases. In fact, in most cases, the trend of lower S02 con-
centrations observed prior to the energy shortage continued. Note, however,
that the emissions and regulatory analysis and the diffusion modeling study
showed that only a limited number of the granted variances were fully im-
plemented throughout the AQCR. These analyses further indicate that if full
implementation had occurred, significantly higher S02 levels would have been
observed across the region.
6. Spatial variations of S02 levels across the region were shown
by the modeling analysis to be primarily a consequence of changes in wind
directional frequencies between 1973 and 1974. The 1974 first-quarter
wind rose indicates a greater frequency of northeasterly flow than occurred
in 1974. This resulted in coastal stations observing lower S02 concentra-
tions during 1973 due to the lack of emission sources to the northeast of
Boston. Receptors to the southwest of Boston observed higher than normal
transport of emissions from the city during that year. Consequently, with
the return of the normal strong northwesterly flow in the winter of 1974,
the total effect of changes in meteorology was to increase S02 levels at
coastal stations an average of 15 percent, which significantly differs from
the 1 percent average increase observed for the AQCR. A decrease in S02
concentrations of 32 percent in the area southwest of Boston, two-thirds
of which is attributable to meteorology, was also observed in 1974.
7. The modeling analysis also indicates that, in the absence of
the energy shortage, TSP levels would have decreased an average of approxi-
mately 18 percent across the AQCR between the first quarter of 1973 and the
/Maiden,
-------
first quarter of 1974, due to the combined effect of growth and changes in
meteorology. The analysis further indicates that the actual decrease
across the AQCR in TSP concentrations was approximately 20 percent, reflect-
ing a combined 2 percent decrease due to the energy-shortage-related para-
meters of conservation (-3 percent) and actual variances implemented (+1
percent). As indicated above, full variance implementation would not have
significantly affected these findings.
The modeling analysis also indicates that the only variance
granted which significantly affected particulate emission rates among fuel
combustion sources in the Metropolitan Boston AQCR was the allowance for
burning coal in three units of New England Power Company's Salem Harbor
power plant. The coastal location and emission characteristics of this
source, coupled with the dominant northwesterly wind flow in the winter
quarter of 1974, resulted in predicted inland concentrations approximately
only 2 percent higher than the hypothetical no-variance case, even though
the facility's particulate emissions increased by a factor of more than 9.
8. Other findings of this program include the indication that
major SIP fuel use regulations effective July 1, 1970, and October 1, 1971
(see Table 2-1), are associated with statistically significant decreases
in regional S02 and TSP levels. This is explained by the fact that both
of these regulations were responsible for conversion of many large fuel
utilization facilities to fuels with lower sulfur and ash contents.
UhUeni
-------
II. AIR QUALITY ANALYSIS
A. BACKGROUND
1. Regulatory
Effective July 1, 1970, the Commonwealth of Massachusetts
Department of Public Health divided Massachusetts into six Air Pollution
Control Districts (APCDs) as shown in Figure 2-1. The Metropolitan Boston
APCD (MBAPCD) was formed from 102 cities and towns comprising 1,400 square
miles and having a population of approximately three million people [2].
Also effective on July 1, 1970, was a ban on all open burning in the MBAPCD
and a limit on the ash content of any fossil fuel burned in the MBAPCD to
nine percent by dry weight. On October 1, 1970 and on October 1, 1971,
several regulations promulgated by the State Bureau of Air Quality Control
(BAQC) limiting the sulfur content of fuels burned in the MBAPCD became ef-
fective. Table 2-1 summarizes these regulations.
On March 9, 1971, the Administrator of the U.S. Environmental
Protection Agency (EPA) designated six Air Quality Control Regions (AQCRs)
in Massachusetts. These six AQCRs are shown in Figure 2-2. Note that three
of these AQCRs are interstate and contain portions of adjacent New England
states while the others correspond exactly to APCDs. The Metropolitan Boston
Intrastate AQCR (MBAQCR) coincides with the Commonwealth's MBAPCD.
On May 31, 1972, pursuant to Section 110 of the Federal Clean
Air Act, the EPA Administrator approved a plan implementing the National
Ambient Air Quality Standards for the Commonwealth of Massachusetts [3].
This State Implementation Plan (SIP) contained regulations for the control
of air pollution adopted to become effective June 1, 1972, under the pro-
visions of Section 142D, Chapter 111, General Laws of the Commonwealth.
All state regulations that were already in effect prior to June 1, 1972
pertaining to control of air pollution were consolidated into the SIP reg-
ulations. Regulations limiting the sulfur and ash content of fuels burned
in the MBAQCR were consolidated into SIP Regulation 5 with a few additions
(see Table 2-1). A complete summary of the SIP regulations is shown in
Table 2-2.
2-1
UJalden
-------
Pioneer Valley
CD
Merrimack Valley
* i'U--'.^
CONNECTICUT
THE COMMONWEALTH OF MASSACHUSETTS
DEPARTMENT OF PUBLIC HEALTH
DIVISION OF ENVIRONMENTAL HEALTM
BUREAU OF AIR USE MANAGEMENT
air pollution control districts
Metropolitan Boston
Central Massachusetts
NOTE NORFOLK COUNTY INCLUDES
8ROOK.LINF. AND COHA55C1
Southeastern ' >£i
Massachusetts I
SCALE IN MILES
7('lo«
-------
TABLE 2-1
A SUMMARY OF REGULATIONS ADOPTED TO CONTROL THE SULFUR AND ASH CONTENT
OF FUELS BURNED IN THE METROPOLITAN BOSTON AIR QUALITY CONTROL REGION*
Effective Date of Adoption
Regulations in Effect in Metropolitan Boston AQCR
i
CO
July 1, 1970
October 1, 1970
October 1, 1971
June 1, 1972
Ban on all open burning
Ash content of fossil fuels limited to 9% by dry weight
Residual oil sulfur content limited to 2.2% and to 1.0% in core**
Distillate oil sulfur content limited to 0.3%
Coal sulfur content limited to 1.5% and to 0.75% in core**
Residual oil sulfur content limited to 1.0% and to 0.5% in core**
Coal sulfur content limited to 0.7% and to 0.37% in core**
Ban on residual oil use at facilities with rated boiler capacities
of 3 million Btu/hour or less
Ban on solid fuel use at hand-fired facilities with rated boiler
capacities in excess of 150 thousand Btu/hour
**
Coincident with the Metropolitan Boston Air Pollution Control District.
The 13 core cities and towns are: Arlington, Belmont, Boston, Brookline, Cambridge, Chelsea,
Everett, Maiden, Medford, Newton, Somerville, Waltham, and Watertown.
-------
Figure 2-2
Air Quality Control Regions in Massachusetts
I
-pi
BERKSHIRE
INTRASTATE.
CENTRAL
MASSACHUSETTS
INTRASTATE
MERRIMACK VALLEY-
iOUTHERN
NEW HAMPSHIRE
INTERSTATE
(MASSACHUSETTS-
NEW HAMPSHIRE)
METROPOLITAN
BOSTON
INTRASTATE
HARTFORD-
NEW HAVEN-
SPRINGFIELD
INTERSTATE
(CONNECTICUT-
MASSACHUSETTS)
METROPOLITAN'
PROVIDENCE
INTERSTATE
(RHODE ISLAND-
MASSACHUSETTS)
-------
TABLE 2-2
SUMMARY OF REGULATIONS FOR THE CONTROL
OF AIR POLLUTION IN MASSACHUSETTS [4]
(Regulations effective June 1, 1972
with amendments effective September 1, 1972)
Regulation
Number Topic
1 General prohibition of air pollution
2 Plant approval and emission limitations (Note: Sections
2.1 and 2.5 amended September 1, 1972)
2.1 New and modified facilities; Department approval
needed for the plans, specifications, standard
operating procedure and maintenance procedures.
2.2 Role of Department in the evaluation of plans
is defined.
2.3 Field of application for Regulation 2 includes
fossil fuel utilization facilities within capa-
cities greater than 3 million Btu per hour, manu-
facturing facilities of all types and dry clean-
ing establishments.
2.4 Department may request and review plans on any
facility which has a likelihood of causing air
pollution.
2.5 Emissions limits specified. Three categories:
(1) existing sources, critical areas of concern
(i.e., urban areas); (2) existing sources, other
areas; and (3) new sources.
Participates: detailed specifications, all types
of facilities.
N02: total emission weight and stack gas concen-
tration limits.
S02: same as N02-
Organic material: requirement for vapor traps.
3 Nuclear energy facilities
4 Fossil fuel utilization facilities (Note: amended
September 1, 1972)
4.1 Plans approved required for facilities with
capacities greater than 3 million Btu per hour
(already required by Regulation 2).
2-5
Maiden
-------
TABLE 2-2
(CONTINUED)
Regulation
Number Topic
4.2 Smoke density indicators required by Apri1 1,
1973 on all facilities with capacities greater
than 10 million Btu per hour (Note: devices
must be approved by the Department).
4.3 All facilities (i.e., existing facilities as
well as new facilities) with capacities greater
than 10 million Btu per hour must have standard
operating procedures approved by the Department
by July 1, 1973.
4.4,4.5 Illegal to remove or circumvent the operation of
control devices.
5 Fuel regulations
5.1 Sulfur content of fuels
5.1.1 Residual oil: Boston core* area: 0.5% S maximum
5.1.2 Remainder of state: 1.0% S maximum
5.1.3 Distillate oil (throughout the state): 0.3% S
maximum. Fuel users may submit alternative plans
for review by the Department. Fuel suppliers must
register with the Department.
5.2.1 No facility with capacity less than 3 million Btu
per hour may use residual oil.
5.2.2 After July 1, 1973. in core* areas, no facility
with capacity less than 6 million Btu per hour
may use residual oil.**
5.3 No solid fuel use permitted in hand-fired faci-
lities with capacities larger than 150,000 Btu
per hour.
*
See Table 1 for definition of core area.
**
This regulation was never implemented by the BAQC due to a procedural
error in which it was omitted from public hearings on the proposed
regulations. The BAQC does not currently plan to implement 5.2.2 in
the future since S02 and TSP standards have been achieved in the core.
2-6
UJalden
-------
TABLE 2-2
(CONTINUED)
Regulation
Number Topic
5.4 Ash content of fuels not in excess of 9%. Fuel
shippers must provide evidence.
5.5 Fuel additives controlled.
5.6 Fuel suppliers must make analyses and maintain
records.
6 Visible emissions (Note: amended September 1, 1972).
Smoke emissions not greater than #1, except for 6 minutes
per hour (then limited to #2). Incinerator emissions
never greater than #1.
7 Open burning prohibited (specific exceptions permitted).
8 Incinerators. Sale of new incinerators and operation of
all incinerators must be approved by the Department.
9 Dust and odor control (Note: amended September 1, 1972).
General prohibition, and specific prohibitions for trans-
portation, construction and demolition.
10 Noise control
11 Transportation media
12 Registration, recordkeeping and reporting. All sources
with energy capacities in excess of 3 million Btu per hour
shall register with the Department annually, and shall re-
port standard data. All other manufacturing sources having
emissions above stated levels shall also register annually.
The Department will acknowledge the registration in writing.
13 Stack testing. Required of all facilities designated by the
Department (present designation: all facilities with capa-
city in excess of 100 million Btu per hour). The operator
of the facility must provide access to stack, etc.
14 Monitoring devices. The Department may require the use of
. emission monitoring devices.
15 Asbestos (added by amendment, September 1, 1972). Spray
application of asbestos fibers is prohibited.
2-7
Maiden
-------
TABLE 2-2
(CONTINUED)
Regulation
Number Topic
50 Variances. (Variances can be granted for periods up to
one year, and must be approved both by EPA and by the
Department. Four items must be covered in the request
for a variance: (1) undue hardship; (2) public good;
(3) evidence that standards will be met; (4) evidence
that no significant deterioration of air quality will
result.)
51 Hearings. Rules for hearings relative to orders and
approvals.
52 Enforcement provisions. Designates building officials,
health, police and fire departments as having authority
in specific cases.
2-8
Maiden
-------
During the summer of 1973, the major oil companies issued
warnings and predictions in relation to shortages of petroleum products.
The events that followed, which came to be known as "the energy crisis",
set in motion the machinery for granting special variances on the Mass-
achusetts SIP air pollution control regulations.
On August 13, 1973, the Governor of the Commonwealth of
Massachusetts asked the Secretary of the Executive Office of Consumer Af-
fairs (EOCA) to conduct a survey and analysis of the heating oil situation
confronting the Commonwealth for the coming winter [4]. Starting in the
summer and throughout the fall, oil companies (the majors and also Mass-
achusetts independents) provided forecasts of distillate oil shortages
to the State Bureau of Air Quality Control (BAQC). Oil retailers, sup-
pliers, and consumers (including industries, utilities, municipalities,
and public authorities) informed State officials and legislators, the
BAQC, and the Governor's Energy Task Force (headed by the Secretary of
the EOCA) of their lack of firm contracts for conforming distillate fuel,
i.e., having a sulfur content less than or equal to 0.3%. Based on their
assessment of a potential oil shortage, the BAQC held public hearings on
September 19, 1973, concerning a proposed relaxation of SIP regulatory
controls on the sulfur content of distillate oil from 0.3% to 0.5% on a
statewide basis. On November 2, 1973, the Massachusetts Department of
Public Health Council voted to approve the relaxation of distillate oil
sulfur content regulations as proposed by the BAQC. On December 11, 1973
[5], the EPA Administrator approved a change in Massachusetts Regulation
5.1.3 which relaxed the sulfur content of distillate oil burned statewide
from 0.3% to 0.5% for the period November 15, 1973 to May 15, 1974.
In addition to distillate fuel, there were also shortages
of low sulfur residual oil. To deal with this situation, the BAQC granted
variances to sources burning residual oil on a case-by-case basis using
three strategies:
• SIP Regulation 5.1.2(d) for large sources
• Additional special variances for large sources
• Special variances for small sources
2-9
Maiden
-------
Regulation 5.1.2(d) allowed statewide use of 2.2% sulfur residual oil by
facilities with a rated boiler capacity at or above 250 million Btu per
hour until May 15, 1974, if certain conditions were met. These conditions
and the full text of 5.1.2(d) are listed in Table 2-3. This modification
to the SIP regulations was approved by the EPA Administrator on January 30,
1974 [6]. On May 16, 1974, EPA published in the Federal Register [7] the
approval for 23 specific sources in Massachusetts to use 2.2% residual oil.
The subset of those sources in the MBAQCR granted this special variance
under SIP Regulation 5.1.2(d) is shown in Table 2-4, along with the effec-
tive dates of the variances. Several other special variances not covered
by Regulation 5.1.2(d) were granted for large sources in the MBAQCR by the
Massachusetts BAQC and are summarized in Table 2-5.
For sources with a rated boiler capacity of less than 250
million Btu per hour, application for a special variance to burn noncon-
forming residual oil had to be received from both the source and the
fuel supplier along with documentation of real need for the variance [10].
Most major fuel suppliers refused to be the principal involved in variance
requests. Instead, they applied intense pressure on their customers to
apply to the BAQC for variances. In all, 601 small-source variances were
granted in the MBAQCR through 12 fuel distributors. These data are sum-
marized in Table 2-6 by supplier. Of the major distributors, only one
independent distributor voluntarily cooperated with the BAQC in redistri-
bution plans to insure priority allocation of the lowest sulfur fuel
available to the Boston core area.
An analysis of the actual quantities of nonconforming fuel
burned in the Metropolitan Boston AQCR under variances granted on fuel
use regulations is presented in Section III.
2. Measured Data
a. Air Quality
Monthly summaries of air quality measurements in the Metro-
politan Boston AQCR for S02 and TSP between January 1970 and June 1974
2~10 ai u
Ulaldeni
-------
TABLE 2-3
Emergency Regulation Adopted by Massachusetts Department of Public
Health on November"2, 1973 (Sulfur in fuel")
5.1.2(d) In the Commonwealth of Massachusetts it vill be permissible for
fossil fuel facilities with an input capacity, as rated by the Department,
at or in excess of 250 million B.t.u./hour to burn residual fuel oil with
a sulfur content up to 2.2£ (1.22 pounds per million B.t.u. heat release
potential) during the period beginning November 15, 1973 and ending May 15,
1971* providing:
a. such facility conforms with guidelines developed by the
Division of Environmental Health,
b. such facility acquires approval in writing from the Department
to operate, and
c. the Division of Environmental Health will conduct a monthly
review of fuel supplies and Air Quality Conditions and; make
appropriate recommendations for amendments to the Public
Health Council.
All such fossil fuel facilities located in the Berkshire Air Pollution
Control District, the Pioneer Valley Air Pollution Control District, the
Central Massachusetts Air Pollution Control District, the Southeastern
I-?assachusetts Air Pollution Control District, the "'errinack Valley Air Pollution
Control District, and the Metropolitan Boston Air Pollution Control District
except those cities and tovns specified in Regulation 5-1.1, * burning a
residual fuel oil with a sulfur content up to 2.25 must have available for
conversion vithin 6 hours of notice from the Department a three (3) day supply
of 1.0$ sulfur content fuel. Notice for such conversion will be given by the
Department in the event of predictions of adverse meteorological conditions.
In the cities and towns specified in Regulation 5.1.1 of the Metropolitan
Boston Air Pollution Control District fossil fuel facilities rated with a
capacity at or in excess of 250 million B.t.u./hour burning a residual fuel oil
with a sulfur content up to 2.2^ must have available for conversion within
6 hours of notice from the Department a three (3) day supplv of 0.55 sulfur
content fuel. Notice for such conversion will be given by the Department in
the event of predictions of adverse meteorological conditions.
* Arlington, Belmont, Boston, Brookline, Cambridge, Chelsea, Everett,
Maiden, Medford, Newton, Somerville, Waltham and Watertown.
2-11
-------
TABLE 2-4
FUEL UTILIZATION SOURCES IN THE METROPOLITAN BOSTON AQCR
WITH A RATED BOILER CAPACITY EQUAL TO OR EXCEEDING 250 MILLION BTU/HOUR
GRANTED A SPECIAL VARIANCE TO BURN 2.2% SULFUR RESIDUAL OIL
UNDER SIP .REGULATION 5.1.2(d) [7]
Source
Location
Eff. Date
Boston Edison Co.
Brandeis University
Cambridge Electric
Mass. Bay Transportation Authority
Mass. Institute of Technology
Revere Sugar Refinery
A.C. Lawrence Leather Co.
Eastman Gelatine Corp.
U.S.M. Corp.*
General Electric Lynn River Works*
Weymouth
Waltham
Cambridge
Boston
Cambridge
Charlestown
Peabody
Peabody
Beverly
Lynn
Nov. 20,
Nov. 30,
Nov. 21,
Nov. 20,
Nov. 21,
Nov. 21,
Dec. 19,
Nov. 20,
Nov. 20,
Nov. 20,
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
2-12
UJalden
-------
TABLE 2-5
ADDITIONAL SPECIAL VARIANCES GRANTED LARGE SOURCES
IN THE METROPOLITAN BOSTON AQCR
Source
Variance
Effective Dates
EPA Approval
ro
i
New England Power Co.*
Salem Harbor Station, Salem
All Units
Units 1,2,3
Boston Housing Authority
13 Housing Facilities in Boston
Boston Edison Co.
New Boston & L St. Stations, Boston
Edgar Station, Weymouth
Mystic Station, Everett
2.6% Sulfur Residual Oil
2.5% Sulfur & 15% Ash Coal
1.0% Sulfur Residual Oil
2.6% Sulfur Residual Oil
Jan. 1, 1974-May 15, 1974
Jan. 23, 1974-May 15, 1974
Nov. 15, 1973-May 15, 1974
Jan. 22, 1974-May 15, 1974
[7]
[8]
[9]
**
NEPCO's Brayton Point Station in Somerset, Mass., was also granted variances but is located in the
Metropolitan Providence Interstate AQCR.
EPA approval not published in the Federal Register.
-------
TABLE 2-6
FUEL UTILIZATION SOURCES IN THE METROPOLITAN BOSTON AQCR
WITH A RATED BOILER CAPACITY OF LESS THAN 250 MILLION BTU/HOUR
GRANTED A SPECIAL VARIANCE TO BURN NONCONFORMING RESIDUAL OIL
No
Fuel Distributor
C.K. Smith & Co., Inc.
Gibbs Oil Co.
Glendale Mortion Petroleum Corp.
(supplied by Union Petroleum)
H.N. Hartwell & Son, Inc.
Mass. Oil & Fuel Co. , Inc.
Metropolitan Petroleum Co.
Northeast Petroleum Corp.
Oil Service of New England
Old Colony Oil Heat, Inc.
Pen-Tex Oil Co.
Union Petroleum Corp.
White Fuel Co.
. of Sources Affected in MBAQCR
In Core* Outside Core Content
3
120
33
--
--
6
1
15
1
224
--
17
--
40
--
7
—
29
--
—
2
2
--
—
--
--
—
1
--
4
--
90
--
6
2.
1.
1.
2.
2.
1.
1.
1.
1.
1.
2.
1.
2.
1.
2.
1.
2.
Effective
Allowed Date**
4%
0%
0%
0%
2%
0%
0%
0%
0%
Q%
2%
Q%
Q%
0%
0%
0%
2%
Jan.
Dec.
Dec.
Dec.
Dec.
Dec.
Dec.
Nov.
Dec.
Nov.
Nov.
Dec.
Dec.
Dec.
Dec.
Nov.
Nov.
10,
7,
28,
28,
11,
13,
11,
30,
26,
30,
30,
11,
28,
13,
12,
30,
30,
1974
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
1973
Refer to Table 2-1 for a list of the 13 core cities and towns.
-------
were obtained through the EPA air quality data handling system, SAROAD,*
and from the BAQC. All data were edited to facilitate completion of a
suitable data base for analysis.
The data base contained several substantive data gaps,
thereby providing an inadequate analysis base at several monitoring sites.
To insure a meaningful analysis, a data subset, consisting of data from
air quality monitoring sites for which there were sufficient measurements
taken, was compiled by excluding any site that had more than two consecu-
tive months of missing data. One exception to this guideline was made
for SCL measurements collected at Kenmore Square in the City of Boston.
Here a single three-month stretch of missing data from March through May
1971 was not used to invalidate the site, as this is the prime benchmark
site used in the development of the State Implementation Plan designed to
insure the attainment of ambient air quality standards in Metropolitan
Boston by 1977 [11]. The final results of the selection process are shown
in Table 2-7 and indicate that SOp data from 13 sites and TSP data from 7
sites are sufficient for data analysis from January 1971 to June 1974. For
the more extended period beginning in January 1970, only 3 sites have suf-
ficient SOo data, and one site sufficient TSP data. Fortunately, however,
the Kenmore Square site is included in both of these sets. The geographical
locations of the selected air quality monitoring sites are shown in Figure
2-3. It should be noted that several of the monthly average concentrations
are based on as few as three daily measurements.
* Storage and Retrieval of Aerometric Data. These published data, main-
tained in the National Aerometric Data Bank by EPA, formed the basis for
the air quality data analysis.
2-15 /7
llUalden/
-------
TABLE 2-7
METROPOLITAN BOSTON AIR QUALITY MONITORING SITES
RECORDING SUFFICIENT DATA FOR TIME SERIES ANALYSIS
IN THE INDICATED TIME PERIODS
SAROAD*
Site Code
Site
Location
Description
Series
Name
Time Span of Data
1970-74 1971-74
Urban
Core
Sites
S02
0240001(F01)
0240002(F01)
0240012
0240013
0340001
• 1160001
S 1200001
1220002
1480002
1700001
1880001
2340003
2620002
TSP
0240002(F01)
0340001
1200001
1480002
1700001
1880001
2340003
Government Center, Boston
Kenmore Square, Boston
South Bay, Boston
Central Square, East Boston
Greenough St., Brookline
Village St., Marblehead
US Army Site, Maynard
Main St., Medford
Dedham Ave., Needham
Nahatan St., Norwood
Hancock St., Quincy
Beaver St., Waltham
Montvale Ave., Woburn
Kenmore Square, Boston
Greenough St., Brookline
US Army Site, Maynard
Dedham Ave., Needham
Nahatan St., Norwood
Hancock St., Quincy
Beaver St., Waltham
Center City-Commercial S02A
Center City-Commercial S02B
Suburban-Industrial S02C
Center City-Commercial S02D
Center City-Residential S02E
Center City-Commercial S02F
Rural-Agricultural S02G
Center City-Commercial S02H
Center City-Residential S02I
Center City-Residential S02J
Center City-Commercial S02K
Rural Agricultural S02L
Center City-Commercial S02M
Center City-Commercial TSP1
Center City-Residential TSP2
Rural-Agri cultural TSP3
Center City-Residential TSP4
Center City-Residential TSP5
Center City-Commercial TSP6
Rural-Agricultural TSP7
Storage and Retrieval of Aerometric Data
-------
Figure 2-3 -,
Air Quality Monitoring Sites with Sufficient Data
for Time Series Analysis
"O,
•--<
STOW
/ ACTCJ
>X
N /
' CONCORD
/ ^T^
/ X
BEDFORD
'/ t
ILEXINGTO
>v
IPSWICH
HAM I Lf Ohi
WENHAM
/DANVERS
PEABOOf
BEVERLY
-
\.
XI
kSALEMl
F<*,
/- LIMCjOLN
r*4
kHANT
sud
BURY
'/ #'
f .T IWE
rwAL.
TON''
.HAM^
J&*1
:?k >r
^
'^h.\
$N
^
<^
X
k/
MNTHRi
X
t
NEWTOhJ
NEE^'AM
53
j"' ^
BOSTON
^^^c
SSHERBORN;
x ">
/> \ c1--
HOLLISrON*^ ^ A
tfOVER
/ \D£
/
NAM"
\.
'MILTON
I . /WESTWOJ
» / /• * '
=>/-^ x' BV
* /
-------
For the selected sites, small gaps of one or two con-
secutive missing data points in the time series of the monthly average
pollutant concentrations were initially filled using a simple linear
interpolation scheme between adjacent months. After preliminary data
analysis by the ratio to moving average technique, a number of the in-
serted values were found to contribute to unusually large irregularities
in the analyzed trends. Using a standard procedure, these were replaced
by more reasonable values suggested by the analysis that incorporated
measurements from similar time periods in different years. The replace-
ment values were obtained by removing the irregular component from the
data time series using the ratio to moving average technique described
in Section II.B.2. With these changes, the data base of S02 and TSP am-
bient air quality measurements, for use in the data analyses, was complete.
A continued review of the individual data points used in
the computation of the discrete time series of monthly SCL and TSP averages
led to selection of three representative series for S02 and TSP each (see
Table 2-8). These six series were used to help sort out and highlight
significant cause-effect relationships in the measured air quality data.
b. Meteorological
Monthly degree-day and wind-speed data were obtained for
Boston for the period January 1970 through August 1974 from Local Clima-
tological Data (LCD) summaries. These data are based on measurements taken
at Boston's Logan International Airport and were used in the data and sta-
tistical inference analyses.
B. DATA ANALYSIS
1. Objectives and Scope
The objective of the data analysis task was to isolate, iden-
tify, and quantify trends in ambient S02 and TSP concentrations in the
Metropolitan Boston AQCR for the period January 1970 through June 1974.
2-18
Maiden
-------
TABLE 2-8
REPRESENTATIVE METROPOLITAN BOSTON AIR QUALITY MONITORING SITES
CHOSEN FOR DATA ANALYSIS
ro
i
SAROAD*
Site Code
Location
Site
Description
Series
Name
Time Span
of Data
SO
0240002(F01)
1700001
2340003
TSP
0240002(F01)
1700001
2340003
Kenmore Square, Boston
Nahatan St., Norwood
Beaver St., Waltham
Kenmore Square, Boston
Nahatan St., Norwood
Beaver St., Waltham
Center City-Commercial S02B
Center City-Residential S02J
Rural-Agri cultural S02L
Center City-Commercial TSP1
Center City-Residential TSP5
Rural-Agricultural TSP7
January 1970-May 1974
January 1971-June 1974
March 1971-June 1974
January 1970-June 1974
January 1971-June 1974
March 1971-June 1974
Storage And Retrieval of Aerometrlc Data
-------
2. Methodology
The data analysis was exploratory in that it examined the
air quality data for trends (or changes in trends) without use of the
knowledge of when fuel use regulations and variances (granted because of
the energy shortage) occurred. In addition, the data were analyzed to
extract, model and quantify persistent periodic behavior related to tem-
poral factors such as seasonal climatology.
Air monitoring data, for ambient concentrations of S02 (13
sites) and TSP (7 sites) in the AQCR, was judged sufficient for data an-
alysis. Time series of monthly pollutant concentrations measured at these
sites were decomposed into three components or movements using a ratio to
moving average technique [12]. These movements are: a trend-cycle com-
ponent composed of both long- and short-term trends in the data, a seasonal
component that quantifies the persistent periodic behavior in the series
that can be related to temporal factors, and an irregular component that
quantifies the non-periodic and unpredictable part of the series. This
analysis was performed using both an additive and multiplicative model of
the series components. Triple exponential smoothing techniques were used
to confirm the trend-cycle component derived in the ratio to moving average
analysis.
3. Results
Although the series components derived using additive and
multiplicative models were comparable in each case, the additive model
produced the smaller irregular component when applied to measured S02 data,
i.e., it was able to explain more of the series behavior in the trend-cycle
and seasonal components. By contrast, the multiplicative model was found
most appropriate for the measured TSP data. Trend-cycle components for
the data series revealed that, in general, regional S02 levels fell con-
sistently during the entire time period of analysis, with the steepest de-
cline occurring prior to 1972 (see Figure 2-4). Ambient TSP levels also
exhibited a general downward trend similar to that of S02 levels, but not
2-20 UbUm
-------
100
E
CD
I c
[N> o
5 50
c
OJ
0
c:
o
CJ
(XI
o
Figure 2-4
Composite Annual S02 Trend-Cycle Components
in Metropolitan Boston AQCR
!llr iiii 'il!R ffi iH
Ambient Air Quality Standard is 80 yg/m . ;
1970
1974
-------
as pronounced or consistent (see Figure 2-5). Figures 2-4 and 2-5 show,
respectively, composite annual average SOg and TSP trend-cycle components
from the multiplicative model in the Metropolitan Boston AQCR with data
from sites inside/outside the urban core area of Boston graphed separately.
Site descriptions are shown in Table 2-7. The procedure followed to pro-
duce Figures 2-4 and 2-5 involved application of the ratio to moving average
technique to the monthly average SOp and TSP measurements taken in Metro-
politan Boston. The multiplicative form of the series components was used,
and seasonal and non-periodic elements of the data time series were filtered
out to obtain the trend-cycle component. Because the seasonal and irregular
components are temporal functions with a mean value of 1.0 in the multipli-
cative model, the magnitude of the trend-cycle component is directly com-
parable to that of the original time series. The monthly trend-cycle com-
ponents were then averaged to obtain annual values, and composite components
were formed by averaging together annual values from different geographical
locations.
A separate examination of annual Boston heating degree-day sta-
tistics (a measure of local fuel burning activities) for the years 1970 through
1974 showed that a long-term trend in meteorological conditions over the time
period of analysis towards lower annual degree-day totals (i.e., warmer win-
ters) did not exist. Consequently, the observed long-term changes in ambient
S02 and TSP levels cannot be attributed to corresponding changes in this
meteorological parameter.
Statistically consistent seasonal variations in pollutant con-
centrations were found in the majority of the data series. Ambient TSP con-
centrations at urban sites displayed seasonal patterns characterized by high
winter values and relatively low summer values. By contrast, seasonal pat-
terns of non-urban sites were typified by high summer peaks and low winter
values. These results parallel the findings of an earlier study [13] of
nationwide TSP levels. Ambient SO^ concentrations were, in general, charac-
terized by high winter values and lower summer values at all sites.
Statistically significant interdependence was found between heat-
ing degree days and wind speed for Boston, reflecting the climatological
2-22 /~~~
•Maiden
-------
Figure 2-5
Composite Annual TSP Trend-Cycle Components
in Metropolitan Boston AQCR
E
CT>
tv>
c
O)
a
Q-
oo
200
±Lli±J rll-U-iUill! i b t i: thltttLttt
Primary Ambient Air Quality Standard is 75 yg/m.
1970
1971
1972
Year
1973
1974
-------
relationship between these parameters. Similar interdependence was found
between these meteorological variables and the data series' seasonal com-
ponents. The implication of this finding is that the meteorological vari-
ables, degree-days and mean wind speed, can be used as a measure of the sea-
sonal ity in ambient SCL and TSP levels. A more detailed summary of the
methodology and results is given in Appendix A (Section VI.A).
C. STATISTICAL INFERENCE
1. Objectives and Scope
The objective of the statistical inference task was to inter-
pret trends (or changes in trends) in ambient S02 and TSP concentrations in
the Metropolitan Boston AQCR for the period January 1970 through June 1974.
The principal focus of the analysis was to test statistically the relation
between trends in regional air quality levels and the implementation of the
SIP fuel use regulations and special variances granted because of the energy
shortage.
2. Methodology
The statistical analysis was confirmatory in that it verified
the existence of trends in the data coincident with the implementation of
fuel use regulations and variances, and it interpreted these trends using
information concerning the regulations and variances. These trends were con-
sistent with those identified in the data analysis task. Nonregulatory fac-
tors simultaneously influencing the measurements, such as meteorological con-
ditions (e.g., heating degree-days and mean wind speed) were also evaluated.
The intercorrelation structure of the principal variables thought to affect
air quality levels with the pollutant concentration series was quantified,
and a ranking of the variables in their ability to explain the meaured levels
of pollutants was produced.
The first step in the statistical analysis involved a procedure
employing the forecast potential of triple exponential smoothing. The tech-
nique used to test the effect of regulations and variances on ambient air qua-
lity levels was to compare statistically the actual measured values after the
2~24 I at u
juJdkjen
-------
implementation date of a regulation or variance with those predicted by the
exponential smoothing model. These latter values were computed under the
assumption of no changes in air quality trends. The analysis was applied
to the seasonally-adjusted form (i.e., seasonal component removed) of the
time series of monthly pollutant concentrations in an attempt to isolate the
effects of only regulations and variances. The results from this analysis
provided the basis for the general multiple linear regression analysis, which
utilized both stepwise and complete regression techniques. The utility of
the multiple regressions was that they permitted identification of which of
a large number of independent varaibles were most significant in explaining
the variations in measured air quality levels. The independent variables
selected for the regression analysis represented meteorological conditions,
seasonal variations, regulatory impositions, and special variances granted.
The dependent variable in the regression analysis consisted of either the
measured ambient SCL and TSP levels or the natural logarithms of these quantities,
3. Results
The results of the statistical analyses indicate that major SIP
fuel use regulations effective July 1, 1970, and October 1, 1971 (see Table
2-1), are associated with statistically significant decreases in regional S02
and TSP levels. This is explained by the fact that both of these regulations
were responsible for the conversion of many large fuel utilization facilities
to fuels with lower sulfur and ash contents. The first set of regulations,
limiting the ash content of fossil fuels burned in the AQCR, caused conver-
sions principally from coal to oil and gas. Note that although this set of
regulations did not limit the sulfur content of fuels, the net results of these
fuel conversions was usage of fuels with a lower ash and sulfur content.
Additional results show that no statistically significant rise in
regional S02 and TSP concentrations occurred coincident with variances granted,
because of the energy shortage, during the winter of 1973-1974. In fact, the
results of the complete regressions indicate statistically significant de-
creases in these pollutant levels occurred regionally commencing in December
1973. The implication of this finding is that some other mechanism, most pro-
bably fuel conservation efforts by consumers, overrode the effects of any in-
crease in SOp and TSP emissions due to the combustion of nonconforming fuel.
2-25
UJalden
-------
The following table indicates the variables and events which,
in the stepwise regression, made the greatest total contribution toward ex-
plaining variations in regional S02 and TSP levels between 1970 and 1974.
These quantities are listed in the order of their ability to explain these
variations.
so2
(Entire Region)
TSP
(Urban Core)
(Non-Urban)
Heating degree-days Heating degree-days (Mean wind speed)
Oct. 1, 1971, regu- July 1, 1970, regu- Oct. 1, 1971, regu-
lations lations lations
In those cases where heating degree-days and the reciprocal of
mean wind speed were significant in the multiple linear regression, they were
found to be directly proportional to measured pollutant concentrations.
Degree-days are a measure of local fuel burning activities and emissions;
mean wind speed is a measure of the dilution capability of the atmosphere.
These proportionality results are consistent with the relationships normally
assumed between pollutant concentrations, emissions, and wind speed in most
air quality models.
The conclusions based on the above results are: (1) fuel burn-
ing emissions dominate S02 concentrations throughout the Metropolitan Boston
AQCR; (2) TSP concentrations in the urban core also are dominated by fuel
burning sources; but (3) TSP levels at non-urban sites are dominated by emis-
sion sources other than fuel burning facilities, most probably local parti -
culate sources, road dust, and pollen. These results confirm the classical
seasonal patterns and the urban/non-urban split found in the measured S02
and TSP data in the data analysis task. A more detailed summary of methodo-
logy and results is given in Appendix A (Section VI.A).
2-26
fchlda
-------
III. REGULATORY AND EMISSIONS ANALYSIS
A. OBJECTIVES AND SCOPE
In order to accurately predict the ambient air quality of a
region through the application of atmospheric diffusion models, it is
essential that an accurate inventory of fuel use and emissions within
that region be available. This inventory must be representative of the
time period of interest and include estimates of the emission rate and
physical characteristics of each emission source. The objective of this
analysis was to create accurate quarterly inventories of fuel use and
emissions in the Metropolitan Boston AQCR for the period of January 1973
to June 1974.
The scope involved in preparing these inventories included up-
dating and projecting a 1972 annual inventory to 1973 and 1974 to account
for:
(1) the effects of heating degree-days on fuel use
(2) the impact of changes in State Implementation Plan
(SIP) regulations on fuel use and emissions
(3) changes in fuel uses induced by population growth,
migration, and economic conditions
and partitioning of these inventories into the quarterly intervals of
interest. The effects of the energy shortage on fuel use and emissions
in each quarter were represented by estimating the degree to which vari-
ance on SIP regulations (granted as a result of the energy shortage) were
implemented and by estimating gross changes in fuel use patterns due to
energy conservation. These quarterly inventories were then used as input
to a diffusion model to predict air quality in the region. Five quarterly
fuel use and emission inventories were prepared for this simulation:
(1) 1973 first quarter emissions reflecting 1973 heating
requirements
(2) A hypothetical emissions inventory reflecting 1973 first
qu-arter emissions adjusted for 1974 heating requirements
(3) 1974 first quarter emissions reflecting 1974 heating
requirements
3-1
UJalden
-------
(4) A hypothetical emissions inventory reflecting 1974 first quar-
ter emissions modified to reflect no variances granted
(5) A hypothetical emissions inventory reflecting 1974 first quar-
ter emissions modified to reflect full variance implementation
The scope involved the following steps:
(1) Acquisition of the 1972 fuel use inventory for the Metropolitan
Boston AQCR from the Bureau of Air Quality Control of the
Commonwealth of Massachusetts
(2) Review of the 1972 inventory for closures and new sources in
• 1973 and 1974
(3) Update of the inventory to 1973 and 1974 using annual degree-
day data to estimate changes in space heating, fuel consump-
tion, and growth factors (demographic indicators) obtained
from the Massachusetts Department of Environmental Affairs
to estimate changes in process fuel use
(4) Modification of the 1973 and 1974 inventories to reflect
gross changes in fuel use patterns and the impact of vari-
ances granted on the State Implementation Plan as a conse-
quence of the energy shortage. Data on the degree of imple-
mentation of the variance were collected by a survey of major
fuel suppliers and users in Metropolitan Boston.
B. METHODOLOGY
This section describes the methodologies used to update an existing
1972 fuel use and emissions inventory to 1973 and the first half of 1974
and to incorporate the effects of the energy shortage into these updated
inventories. This section includes a brief discussion of the inventory
structure, a description of the base year inventory, and a summary of the
assumptions and methods used to project and apportion the inventories.
Appendix B includes a more detailed description of each of these steps.
1. General
All fuel consumption can be broadly categorized into use by
stationary sources and by non-stationary (transportation) sources. This
latter category comprises all modes of combustion-power transportation,
including automobiles, busses, trucks, trains, vessels, and aircraft.
3-2
-------
Stationary sources may also be classified into different user categories
and type of use. The major categories of source types include domestic
(residential), commercial, institutional, manufacturing, and steam-elec-
tric utilities. Most of these can be further classified as point sources,
which represent individual establishments using large quantities of fuel,
as distinguished from area sources, which represent, collectively, a large
number of smaller sources distributed over the survey area. In addition,
certain types of activity which also produce air pollutants, such as the
marketing of gasoline, the use of solvents, on-site incineration, forest
fires, and structure fires are included as area sources.
The methodology used in the analysis required that accurate
data be available for all of these fuel use and emissions categories.
Each category was then analyzed to determine what type of demographic
indicator would best represent annual growth in this category and what
portion of fuel use would have a seasonal .component.
2. Base Case
The base case inventory used for this study was an inventory
in National Emission Data System (NEDS) format reflecting total fuel use
in the Metropolitan Boston AQCR for the calendar year 1972. The point
source section of this inventory is based on information collected by the
Massachusetts Bureau of Air Quality Control (BAQC) during the period 1970
through 1972. The area source section of the inventory was developed by
Wai den Research and the BAQC in support of an Air Quality Maintenance Plan
for Massachusetts by disaggregation of statewide non-point source fuel to
the APCDs by demographic indicators and to a grid system by land use factors
[22].
The 1972 point source inventory contains approximately 400
sources. These data were reviewed on a source-by-source basis to insure
that all SIP regulations in effect during 1972 were reflected in the inven-
tory. No attempt was made to update fuel use rates to 1972 for source
Maiden
-------
information received prior to that year (i.e., 1970-1971) unless the
information conflicted with 1972 SIP regulations. This was done for
three reasons: (1) all large emission sources had been previously up-
dated by the BAQC with 1972 operating data; (2) prior to the energy
shortage, the annual change in fuel use of a particular source was di-
minishingly small; and (3) the area source fuel use inventory was created
by subtracting the fuel use of this point source inventory from the total
fuel use in the AQCR apportioned from data reported by the Bureau of
Mines [23].
The 1972 NEDS-based area source inventory is composed of
2
about 1,700 grid squares ranging in size from one to 256 km , located
throughout the AQCR. Each grid square contains an estimate of fuel use
for 31 source classifications within the area represented by the grid.
2
A number of the smaller grids (less than 64 km ) located more than 15 km
from the Boston central business district were combined to form larger
2
grids, approximately 100 km in area.
Emission factors used for this study are those assembled and
published by the Environmental Protection Agency [24] with the exception
of fuel combustion particulate emissions. The latter factors were de-
rived from a study of the local mix of burner-boiler types throughout the
AQCR [25]. Table 3-1 tabulates the emission factors used in preparing
the inventory. Emission rates were computed by assuming the maximum
emissions allowable by the SIP regulations for each source category, fuel
type, and control device. Exceptions to this rule were several power
plants for which the actual average 1972 sulfur content of the fuel was
provided.
3. Projection and Apportionment Procedures
Projection of the base year inventory to 1973 and 1974 and
apportionment of these projected inventories to quarters was accomplished
by execution of the following tasks.
UlaUen
-------
TABLE 3-1
EMISSION FACTORS
(mass of emissions/fuel unit burned)
Sulfur Dioxide
Metric1" English
Particulates
Metric English
FUEL OIL
Residual Oil
6 *
Sources> 9x10 j/sec
6 *
Sources< 9x10 j/sec
Distillate Oil
NATURAL GAS
COAL
**
19S
19S
17S
9.6
19S
157S
157S
140S
0.6
38S
1.0
1.1
0.3
128
***
6.5A
8
9
140
8
13A
30 ft ?
+ Metric units: Mass - kg; fuel oils - m ; coal - 10 kg; natural gas - 10 m
3 63
++ English units: Mass - Ib; fuel oils - 10 gal; coal - ton; natural gas - 10 ft
* 8.8xl06 j/sec = 30xl06 Btu/hr
** Percent sulfur by weight
*** Percent ash by weight
Illtiden,
-------
(a) Addition of new point sources not included in the
base inventory and deletion of point sources closed
during the period of interest
(b) Adjustment of process fuel use to reflect economic
growth
(c) Adjustment of all space heating fuel use to reflect
annual degree-day differences
(d) Adjustment of area source space heating fuel use
to reflect population growth
(e) Adjustment of the emissions inventory by fuel type
and establishment to reflect the degrees of imple-
mentation of variances granted by EPA to the facility
(f) Adjustment of the fuel use inventory by fuel type
and source category to reflect energy conservation
•(g) Verification of 1973 fuel use projections by fuel
type and source category by comparison of these es-
timates with apportioned published regional totals [23]
(h) Apportionment of the projected inventories to quar-
ters by allocating space heating fuel use to quarters
based on the quarterly variation of degree-days within
the year of interest and by allocating process fuel
used to quarters based on uniform fuel use through-
out the year.
A detailed summary of each of these steps is presented in Appendix B-l.
C. RESULTS
1. Fuel Use Trends
a. Trend Components
Based on historical indicators, the expected growth in
fuel use in the Metropolitan Boston area had been quite small during the
1970's [23]. Between 1970 and 1973, there was no net change in residual
oil use in Massachusetts; distillate oil use increased only one percent
annually. This situation resulted from two factors:
(a) The tendency towards no growth in regional
population restricted any new growth in the
residential and commercial sectors [26]
(b) The high cost of labor and materials in this
area has forced many manufacturing firms out
of New England. Thus, total process emissions
and industrial fuel use were expected to remain
constant or even decline in the 1970's.
Uhlden
-------
These conditions are reflected in the demographic growth factors* obtained
from the Commonwealth of Massachusetts [26] which indicated that average
annual increases in manufacturing employment would be about 0.6%, non-
manufacturing employment about 2.4%, and population about 0.6%, between
1970 and 1975 within the Metropolitan Boston SMSA. The change in popula-
tion by city and town indicated that populations in the core area (Boston
and the 12 surrounding cities and towns) would decrease by 0.3% and that
the outlying areas would increase at a rate of 1.4% annually.
With the onset of the energy shortage, a significant
change in fuel use patterns occurred. The scarcity and high cost of all
fuels resulted in considerable conservation efforts by both private and
public sectors of the economy. The 1974 consumption of distillate oil,
which is used primarily for space heating, was reduced by 15% from 1973
levels after accounting for differences in meteorological heating require-
ments, while smaller users of residual oil showed savings of 12% (see dis-
cussion in Appendix B-2). The electric utilities, which consume approxi-
mately 78% of the residual oil burned in the region, showed a decline in
fuel use for all purposes of about four percent.
The Metropolitan Boston area consumes a considerable amount
of fuel for space heating. Thus, the total heating requirements for a
given year strongly influences total fuel demand. As shown below, the base
year, 1972, was approximately five percent colder than the 30-year clima-
tological average for the region (as measured by degree-days), whereas 1973
was nine percent warmer than average, and 1974 was approximately equal to the
1972 average.
DEGREE-DAY DATA
Logan International Airport
Degree-Days
Year Annual First Quarter
1972 5914 2889
1973 5144 2670
1974 5898 2832
30-Yr Mean 5621 2913
*
See Table B-l in Appendix B.
3-7
Maiden
-------
b. Analysis of Component Effects
The combined effects of growth and conservation on fuel
use indicate that a small increase in the annual rate of fuel use occurred
during the first three quarters of 1973. With the advent of the energy
shortage in the fourth quarter of 1973, major conservation efforts began
and by the first quarter of 1974, completely overshadowed any increases
in fuel use due to economic growth.
The effect of meteorology* on fuel use was to decrease
fuel use in 1973, especially residential and commercial distillate which
are used primarily for space heating. Empirical data indicate that space
heating fuel use is not linearly proportional to degree-days [27]. There-
fore, this fuel use will decrease by less than the 13% differential in
degree-days between 1972 and 1973. Fuel use in 1974 would have changed
little from the 1972 total, if only meteorology was used as an indicator.
Appendix B documents the methods used to determine and
incorporate the effects of growth, conservation, and meteorology. Table
3-2 presents total fuel use by fuel type and source category for the three
years of interest.
2. Regulatory and Variance Analysis
The Bureau of Air Quality Control (BAQC) of the Commonwealth
of Massachusetts has adopted a number of regulations affecting fuel burn-
ing and process emission sources designed to control S02 and particulate
emissions. Table 2-1 presents a summary of the regulations relevant to
this study. Other than a regulation requiring extremely small users of
residual oil to convert to distillate oil, no significant regulations were
enacted during the period of this analysis.
Variances in these regulations were granted in order to al-
low for greater leeway in the quality of fuels burned in anticipation of
a significant shortage of conforming fuel.
*
Degree-days.
Ulalda
-------
TABLE 3-2
ANNUAL FUEL USE IN METROPOLITAN BOSTON
CO
I
Distillate (102m3)
Residential
Commercial -Institutional
Industrial
Residual (102m3)
Electric Generating
Industrial
Commercial -Institutional
Natural Gas (106m3)
Industrial
Commercial -Institutional
Residential
Bureau
1972
25,797
14,095
7,918
47,810
37,325
15,761
21,269
74,355
458
510
1,358
2,326
of Mines*
1973
24,554
13,374
8,501
46,429
36,088
16,187
20,647
72,922
428
545
1,323
2,296
1972
25,797
14,278
7,966
48,042
37,612
15,691
21,514
74,527
457
535
1,356
2,348
Inventory**
1973
24,476
13,895
7,978
46,349
35,967
15,604
20,811
72,382
459
518
1,286
2,263
1974
21,777
11,966
6,983
40,726
33,963
13,689
18,045
65,697
400
448
1,145
1,993
**
Actual totals apportioned to MBAQCR from state total.
Sum of inventory entries.
Note: Coal used by external combustion sources in 1972 was limited to one source (Boston Engine Terminal)
which used 1.32 x 108 kg of bituminous coal. This source converted to residual oil in 1973. Coal
use in 1974 resulted from variances granted to the Salem Harbor Plant. This plant burned 15.7 kg/sec
bituminous coal during the period when the variance was in force.
Note: 102 m3 = 2.64 x 10" gal; 106 m3 = 35.3 x 106 ft3
-------
Variances granted in the Metropolitan Boston AQCR as a
consequence of the energy shortage were of four types:
(a) A blanket relaxation of the sulfur in fuel limita-
tion in distillate fuel from 0.3% to 0.5%
(b) 300-400 individual variances relaxing residual fuel
limiations from 0.5% and 1.0% to up to 2.2%
(c) Conversion of Units 1-3 at the Salem Harbor Power
Plant from oil to coal
(d) Operation of some sources 21 250 MBtu/hr heat input
allowing combustion of residual fuel oil with up
to a 2.6% sulfur content.
A study of the actual implementation of these variances has indicated that,
for the most part, they were only implemented to a limited extent. This
resulted from the shortage of all grades of oil. The variance granted which
was most nearly fully implemented was the allowance for three units of the
Salem Harbor Power Plant to burn coal. This source burned 15.7 kg/sec of
coal containing 15.6% ash and 0.77% sulfur during the entire period of the
variance.* The variance allowed for 15.6% ash and 2.5% sulfur coal.
The use of nonconforming fuels increased SO- emissions in the
AQCR from distillate users by only 2 percent, based on the assumption that
all conforming fuel used during the period was equal to the SIP regulatory
limit. Average residual oil sulfur content from large power plants did not
exceed the SIP regulations, although small residual users in the core area
of the AQCR which were granted variances used oil with an average sulfur
content 13 percent above the regulatory limit. Table 3-3 presents a sum-
mary of the actual and allowable sulfur content for residual oil users
granted variances.
Particulate emissions were generally not affected by the
variances granted; the particulate emission rate for the various fuel types
varies very little with increasing sulfur content. However, the variance
allowing the Salem Harbor Power Plant to burn coal increased average parti-
culate emissions from this plant 900 percent during the first quarter of
1974.
* January 23, 1974, to May 15, 1974
3-10 Ulalden
-------
Table 3-4 summarizes annual emissions of S02 and particu-
lates in the Metropolitan Boston AQCR. Detailed information on variance
implementation is contained in Appendix B-3.
3-11
UJakJen
-------
TABLE 3-3
IMPLEMENTATION OF VARIANCES GRANTED TO
SOURCES BURNING RESIDUAL OIL
Source
Boston Edison
Edgar
Mystic
New Boston
and L Street
New England Power Co.
Salem Harbor Plant
Units l , 2 and 3
Unit 4
Cambridge Electric
Blackstone
and Kendall
MBTA
M.I.T.
Revere Sugar
A. C. Lawrence
Eastman Gelatine
USM Corporation
General Electric
Lynn River Works
Brandeis University
Lipton Pet Foods
SIC
Code
4911
4911
4911
4931
8221
2C62
3111
2891
3999
3562
8221
2010
2042
1972
3
meters
373,150
746,200
1,320,400
524,800
314,260
98,820
62,200
283,000
16,482
45,480
29,020
52,260
1,786
116,410
12,410
2,089
3,346
Fuel Use
103 gals.
98,600
197,160
348,900
138,660
83,040
26,100
16,440
74,780
4,355
12,000
7,668
14,337
472
30,760
3,280
552
884
% Sul
no
variance
0.94
0.49
0.44
0.65
0.65
0.50
0.41
0.50
0.50
0.50
1.00
1.00
1.00
1.00
0.50
1.00
1.00
fur in
actual
imple.
0.95
0.49
0.48
coal
0.75
0.56
0.56
0.56
0.56
0.56
1.00
1.00
1.00
1 .00
0.56
1.00
1.00
Fuel
full
variance
2.2*
2.6**
2.6**
2.6**
2.6**
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.2
2.0
2.0
* Until January 22, 1974, and 2.6% from then until May 15, 1974. Average full
variance sulfur content: 2.5%
** Not approved until January 22, 1974. Average full variance sulfur content: 2.1%
Mot approved until January 22, 1974. Average coal sulfur content: 0.77%
3-12
Illlakk
-------
TABLE 3-3 (cont.)
IMPLEMENTATION OF VARIANCES GRANTED TO
SOURCES BURNING RESIDUAL OIL
Source
Middlesex County
Hospital
New Enqland Confec-
tionary
U.S. Gypsum Company
William Carter Co.
M.M. Mades Company
Columbia Packing
American Bil trite
Rubber Company
Container Corpora-
tion of America
(Medford)
Tufts University
William Underwood Co.
Draper Brothers
Amstar Corporation
Converse Rubber Co.
General Electric
First National
Stores
SIC
Code
8062
2071
3274
2341
2013
2013
3069
2691
8221
2013
2231
2062
3021
3722
5411
1972
3
meters
1,896
4,129
3,410
749
787
2,301
7,350
1,972
6,032
1,124
2,422
18,896
5,313
4,552
5,423
Fuel Use
103 gals.
501
1,091
901
198
208
608
1,942
521
1,594
297
640
4,993
1,404
1,203
1,433
% Sul
No
variance
0.50
0.50
0.50
1.00
0.50
0.50
0.50
0.50
0.50
0.50
1.00
0.50
0.50
0.50
0.50
fur in
actual
imple.
0.56
0.56
0.56
1.00
0.56
0.56
0.56
0.56
0.56
0.56
1.0
0.56
0.56
0.56
0.56
Fuel
full
variancp
2.40*
1.00
1.00
2.20
1.00
1.00
1.00
1.00
1.00
1.00
2.20
1.00
1.00
1.00
1.00
Mot effective until January 10, 1974.
3-13
Maiden
-------
TABLE 3-3(cont.)
IMPLEMENTATION OF VARIANCES GRANTED TO
SOURCES BURNING RESIDUAL OIL
Source
Avco Research Labs.
Pel ton and Sons
Solvent Chemical Co.
National Laundry
Whidden Memorial
Hospital
Northeastern Univer-
sity
Cambridge Thermionic
SIC
Code
7391
2085
2818
7210
8061
8221
3679
1972
Meters
2,566
1,177
787
1,052
715
2,774
146
Fuel Use
103 gals.
678
311
208
278
189
733
552
% Sul
No
variance
0.50
0.50
0.50
0.50
0.50
0.50
0.50
fur in
Actual
imple.
0.56
0.56
0.56
0.56
0.56
0.56
0.56
Fuel
Full
variance
1.00
1.00
1.00
1.00
1.00
1.00
1.00
Corporation
3-14
-------
TABLE 3-4
ANNUAL SULFUR DIOXIDE AND PARTICULATE EMISSION RATES
IN METROPOLITAN BOSTON
(103 metric tons)
Sulfur Dioxide
1972. 1973 1974**
Particulates
1972 1973 1974**
External Combustion Boilers*
Residual Oil
Electrical Generation
Industrial
Commercial -Institutional
Distillate Oil
Industrial
Commercial -Institutional
Residential
Coal
Process Emissions
Incineration
Mobile
TOTALS
43.7
24.8
33.1
4.0
7.2
13.0
0.2
1.3
1.3
3.1
131.7
41.4
24.4
31.9
4.0
6.7
12.3
0.0
1.3
1.2
3.2
126.4
36.0
22.0
28.4
3.6
6.1
11.2
6.4+
1.2
3.1
118.1
3.7
1.7
2.4
0.2
0.4
0.7
1.0
0.2
9.1
9.5
28.9
3.6
1.7
2.3
0.2
0.4
0.7
A n
-------
IV. MODELING ANALYSIS
A. OBJECTIVES AND SCOPE
The objective of the modeling analysis was to isolate and quan-
tify the effect of meteorology, growth and conservation, and source class
variances on sulfur dioxide (S02) and particulate levels in the Metropoli-
tan Boston AQCR during the period of the energy shortage in order to obtain
a clearer understanding of the impact and potential impact of that shortage
on ambient air quality. The scope of this analysis included prediction
of ground-level concentration of S02 and particulates during the first
quarter (January - March) of 1972, 1973, and 1974 at monitoring sites
throughout the AQCR, reconciliation of the predicted concentrations with
the observed data,calibration of the model to reflect the reconciliation,
and extension of the modeling analysis to a series of actual and hypothet-
ical conditions to permit isolation of the effect of the parameters of
interest on air quality during the winter of 1973-1974.
B. METHODOLOGY
This section describes the methodology used to perform the dif-
fusion analysis and the procedures used to disaggregate the change in air
quality between the first quarter of 1973 and 1974 into the components of
interest. Included is a brief description of the diffusion model, a sum-
mary of the validation/calibration process, and a description of the simu-
lation cases which were studied. Appendix C provides a comprehensive des-
cription of these analysis elements.
1. Model Characteristics
The diffusion model used to simulate the various cases was
the Air Quality Display Model (AQDM) [28] in the form of a segment of the
Air Quality Implementation Planning Program (IPP) [29]. This computer
program is a formalization of a simulation model originally developed by
Martin and Tikvart [30] and has received wide application in the evaluation
of regional air,quality.
4-1
-------
The AQDM is designed to provide seasonal or annual predic-
tions of ambient concentration levels. The average concentration at a
receptor point from a given source is determined by solving the diffu-
sion equation for specified combinations of meteorological conditions,
and weighting these by the frequency of occurrence of these conditions.
A total of 480 combinations are considered based on 16 wind direction
azimuths, six wind speed classes, and five stability categories. The
total concentration at the receptor point is given by the summation of
contributions from all other emission sources in the study area.
This model has been modified from its original form to in-
clude the Briggs plume rise estimates and to produce a source contri-
bution file to facilitate case evaluation. Appendix C-l presents more
detailed information of the model, the assumptions involved in executing
the simulations, and on the simulation scaling procedure.
2. Reconciliation of Measured Air Quality Concentrations
with Model Predictions
Using the emission inventories developed in the Regulatory
and Emissions Analysis (Chapter III) which estimate actual conditions for
the first quarters of 1972, 1973 and 1974, predictions of average concen-
trations of sulfur dioxide and particulates were made for monitoring sta-
tions operated by the Massachusetts Department of Public Health within the
Metropolitan Boston AQCR. These predictions were then compared with the
measured data at these sites during the same time period to permit cali-
bration of the diffusion model.
a. Sulfur Dioxide
To provide consistency with the results of the air qual-
ity data analysis (Chapter II) and to eliminate potential inconsistencies
due to instrumental differences, only monitoring sites using the West-Gaeke
bubbler sampling technique were used in the validation process. The data
4-2
-------
from these monitoring stations were carefully screened to assure that a
representative quarterly average was obtained* (see Appendix C-2). This
process resulted in seven sites for validation in 1972, four in 1973,
and eight in 1974. Various regression techniques were studied in order
to develop a single statistically significant calibration which could be
applied to the entire AQCR. The technique selected was a forced zero-
intercept linear regression. This technique was used because the natural
background for S02 is essentially zero.
b. Particulates
A validation procedure similar to that described for St^
was performed at stations measuring total suspended particulates (TSP)
using high-volume samplers within the Metropolitan Boston AQCR. There
were only three monitoring sites with adequate data for validation in each
year.
The small number of validation points considerably limits
the ability to develop a statistically significant calibration procedure
in any one year. Therefore, the calibration was developed using data for
all years combined. The slope-intercept method of calibration was used
for TSP to allow for the natural TSP background.
3. Description of Simulations
Five simulations were undertaken to isolate the relative ef-
fects of three parameters on air quality: meteorology, SIP variances,
and growth and conservation. These simulations were:
(1) 1973 first quarter emissions inventory data with 1973
dispersion data
(2) A hypothetical inventory reflecting 1973 first quarter
emissions inventory patterns modified for 1974 degree-
days, with 1974 dispersion data
* A quarterly average was considered adequate for validation if there were
at least three observations made within each of the three months. This
limited number of observations is used because most non-continuous moni-
toring is performed only once every six days, resulting in a maximum of
four or five observations per month. f~
Wakkn
-------
(3) 1974 first quarter emissions inventory data with 1974
dispersion data
(4) A hypothetical emissions inventory reflecting 1974
first quarter emissions inventory data, modified to
simulate no variances granted, with 1974 dispersion data
(5) A hypothetical emissions inventory reflecting 1974
first quarter emissions inventory data, modified to
simulate full implementation of variances granted, with
1974 dispersion data.
These simulations were performed by selecting various combinations of the
parameters of interest. Comparison of the results between any two simula-
tions isolates the effect of the parameter not common to each. Table 4-1
presents the parameters incorporated in each simulation. The various
choices of parameters are as follows.
a. Fuel Use Patterns - Conservation and Growth
This parameter allows a choice of 1973 fuel use patterns
which include no conservation efforts of 1974 fuel use patterns reflecting
only the change in fuel use between 1973 and 1974 that would result from
growth and conservation efforts.
b. Meteorology
This category allows a change in fuel use dependent only
upon difference in space heating requirements (as measured by degree-days)
between 1973 and 1974. Also, a choice of atmospheric dispersion character-
istics representative of 1973 or 1974 is allowed. These effects were not
specifically separated in this study.
c. SIP Variance
Three conditions are allowed in this category: no variance
implementation, estimated actual variance implementation, and full variance
implementation.
Simulations 1 and 3 represent the actual conditions which
occurred during 1973 and 1974, respectively. The remaining inventories are
hypothetical, i.e., they never actually occurred.
4-4
-------
TABLE 4-1
SIMULATION COMPONENTS
Simulation
2345
Fuel Use Patterns*
Conservation & Growth / / /
Meteorology**
1973 /
1974 / / / /
SIP Variance
None / / /
Actual ,
Full v
*
Assumes 1973 to be the base inventory.
**
Includes both degree-day and dispersion data.
4-5 lUhldenl
-------
The differences among these simulations can be com-
pared to isolate the effects of interest. Meteorological effects are the
only difference between simulations 1 and 2; thus, the difference between
concentrations predicted by these simulations isolates the effects of
meteorology between 1973 and 1974. The only difference between simula-
tions 3 and 4 is the extent to which variances were actually implemented,
thus isolating that effect. Similarly, simulations 2 and 4 isolate the
effects of growth and conservation and simulations 4 and 5 the effects of
full variance implementation.
The Metropolitan Boston AQCR contains a considerable
variety of geographical areas; coastal, urban, suburban, and rural. The
spatial variation in the effects within each simulation can also be studied
by comparing receptors grouped by the various geographical types.
Sulfur dioxide and total suspended particulate concen-
trations were, therefore, predicted at 50 receptor points for each simu-
lation case throughout the Metropolitan Boston AQCR. The base grid used
was a square comprised of 25 prediction points spaced at 10 kilometer in-
tervals on a side. The remaining receptors were located at the coordinates
of the Commonwealth's S02 and TSP monitoring stations in the AQCR, and at
other special interest sites. Figure 4-1 shows the receptor locations.
C. RESULTS
1. Calibration
a. Sulfur Dioxide
As noted previously, the calibration procedure used to
model S02 ambient concentration was a linear regression forcing the in-
tercept of the regression to zero. The following summarizes the results
of this analysis for the first quarter of 1972 and 1973.
4-6
UJaldff
-------
i"x^ ,' v
= Base Grid Receptors
A= Additional Receptors
Figure 4-1. Receptor Locations.
-------
Time Period Correlation Coefficient Line
1972 0.75 Y=0.86X
1973 0.48 Y=0.83X
The extreme scatter of the measured data in 1974 resulted in no correlation
with predictions.
The 1972 correlation coefficient indicated that the re-
gression line is a statistically significant fit. The 1973 regression is
not statistically significant; however, the line of prediction for these
data is statistically identical with that of 1972. The inability to vali-
date the 1974 measured data appears to be the result of non-representative
samples and not errors in the modeling analysis. When plotted, however,
the 1974 data also appear to be well represented by the 1972 line of pre-
diction. Therefore, the 1972 line of prediction was used to calibrate all
simulations. Figure C-l in Appendix C-2 presents a complete plot (1972-
1974) of this validation. Appendix C-2 also presents the data used in this
validation and discusses the calibration attained by using the best-fit
slope-intercept form of validation, which, by definition, gives a better
fit for the data.
b. Particulates
A regression using the slope-intercept technique was
performed with the particulate data. However, due to the limited number
of receptors, a year-by-year calibration was not attempted.
The regression analysis indicates that the line of best
fit is Y = 1.07 X + 20. The correlation coefficient for this line is
0.73, which is statistically significant at the 95% level. Appendix C-2
presents the data used in this validation.
4-8
-------
2. Modeling Results
a. Air Quality Predictions
so2
r
Modeling of the five cases resulted in average first
quarter air quality predictions for the Metropolitan Boston area. Figures
4-2 through 4-6 present the predictions of SOg concentrations for each of
these respective cases. All data on these figures reflect application of
the calibration technique described in the preceding subsection.
The maximum concentration observed in the AQCR for each
case occurs in the central business district of Boston. Concentrations at
this point vary little (less than 10 percent) for each of the first four
cases. Concentrations resulting from the hypothetical full variance case
(Simulation 5), however, were considerably greater than the concentration
predicted in 1973. The only significant change in isopleth patterns among
the cases is the tendency for isopleths to extend to the southwest of Boston
during 1973 and to the southeast of Boston for the cases which simulate 1974
meteorology.
TSP
The principal sources of particulate emissions in the Metro-
politan Boston AQCR were not significantly affected by the energy shortage,
as was noted in Section III. The TSP isopleths for the five cases, Figures
4-7 through 4-11 respectively, verify this observation. These data indicate
that there was no significant change in TSP levels at any receptors predicted
by Simulations 1, 2, and 4. Simulations 3 and 5, the 1974 actual implementa-
tion and hypothetical full variance cases, show slightly increased concentra-
tions in the vicinity of the Salem Harbor Power Plant, which was burning coal
during the first quarter of 1974. Note that the actual implementation at
Salem Harbor was identical with full variance implementation for particulate
matter.
4-9 /7
IIIMen!
-------
I
o
Figure 4-2. Sulfur Dioxide Concentrations (yg/m3) 1973 First Quarter - Simulation
-------
-p.
I
Figure 4-3. Sulfur Dioxide Concentrations (yg/m3)
1973 Fuel Use Patterns - Simulation 2
1974 First Quarter Meteorology and
-------
I
ro
Figure 4-4. Sulfur Dioxide Concentrations (ug/m3)
Imp! ementa-ti on - Simulat-ion 3
1974 First Quarter Actual Variance
-------
I
CO
Figure 4-5. Sulfur Dioxide Concentrations (yg/m3)
Implementation - Simulation 4
1974 First Quarter No Variance
-------
Figure 4-6.
Sulfur Dioxide Concentrations (yg/m3)
Implementation - Simulation 5
1974 First Quarter Full Variance
-------
I
en
Figure 4-7. ISP Concentrations (yg/m3) 1973 First Quarter - Simulation 1
-------
CT>
Figure 4-8. TSP Concentrations .(uq/m3) 1974 First Quarter Meteorology and 1973
Fuel Use Patterns - Simulation 2
-------
Figure 4-9. TSP Concentrations (yg/m3) 1974 First Quarter Actual Variance
Implementation - Simulation 3
-------
I
«_J
00
Figure 4-10. TSP Concentrations (yg/m3) 1974 First Quarter No Variance Implementation
Simulation 4
-------
Figure 4-11. TSP Concentrations (yg/m3) 1974 First Quarter Full Variance Implementation
Simulation 5
-------
b. Isolation of Effects
Modeling of the five cases permitted isolation of the
relative effects of meteorology, variance implementation (actual and hy-
pothetical "worst case"), and the combined effects of growth and conser-
vation on ambient air quality in the AQCR during the energy shortage.
(1) Sulfur Dioxide
Table 4-2 presents the average percentage change in
sulfur dioxide concentrations between the first quarter of 1973 and the
first quarter of 1974 based on model predictions for the base grid (simu-
lations (1) and (3) described in Section IV.B.3). The base grid is 25
receptors equally spaced throughout the AQCR which are assumed to repre-
sent average change.
TABLE 4-2
S00 Component
Net Change Growth and
1973-1974 Weather Conservation
Mean Percent Change -12.0 1.3 -16.0
Standard Deviation 15.8 17.5 3.5
Variances
Implemented
2.7
2.2
Also included in this table is a dissection of the total change into the
three components of interest: meteorology, growth and conservation, and
variances implemented. These results indicate that the most significant
influence on S02 levels in this period was the growth and conservation
parameter.*
Note that "growth" rate was less than 2% for all inventory categories
between 1973 and 1974.
4-20
Waiden
-------
S02 ambient concentration levels improved (i.e., decreased)
by 16 percent as a consequence of growth and conservation, while meteorology
and "variances implemented" tended to degrade air quality very slightly.
Table 4-2 also presents the standard deviation of the model predictions.
These statistics indicate that the observed effects of conservation and
growth, and of "variances implemented" vary little throughout the region.
However, the effects of weather vary markedly; primarily due to differences
in wind directional frequencies between 1973 and 1974.
Table 4-3 presents the average percentage changes in
S02 concentrations that can be attributed to the causal components, as
derived from the simulations for the base grid and various geographically
similar groups of receptors. Table 4-3 also presents an indication of
what would have resulted from full implementation of all variances. These
estimates indicate that, in general, S02 concentrations would have increased
37 percent over the "no variance" situtation.
The variable effects of meteorology, especially due
to changes in wind direction frequency, are evident from Table 4-3. The
differences have been analyzed based on Figure 4-12 which presents the
first quarter wind roses for 1972, 1973 and 1974, based on observations
made at Boston's Logan Airport. The 1973 wind rose indicated a greater
frequency of occurrence of wind flows from the northeast than occurred in
1974. This resulted in coastal stations observing lower S02 concentrations
during 1973 due to the lack of emission sources to the northeast of Boston.
Also, receptors to the southwest of Boston observed higher than normal
transport of emissions from the city during that year. Consequently,
with the return of the more normal dominant northwesterly flow in the
winter of 1974, coastal station S02 levels increased an average of 15%,
which significantly differs from the 12 percent average decrease observed
for the AQCR. A decrease in S02 concentrations of 32 percent in the area
southwest of Boston, two-thirds of which is attributable to meteorology
(i.e., dispersion and degree-days), was also observed in 1974.
4-21 /
Utiden,
-------
TABLE 4-3
PERCENT CHANGE IN SULFUR DIOXIDE CONCENTRATIONS
AS PREDICTED BY MODEL
FIRST QUARTER 1973 TO FIRST QUARTER 1974
BASE RECEPTOR GRID
Receptor
Grouping
Base
Urban Locations
Coastal Locations
Southwest Boston
Net Change
1973 - 1974
-12.0
-13.1
0.8*
-32.0*
Weather
1 .3
0.4
15.3*
-21.1*
Components
Growth and
Conservation
-16.0
-17.2
-19.4
-12.4
Variances
Implemented
2.7
3.6
4.8*
1.5
Full Variance
Implementation
52.3
51.3
71.9
28.1*
Significantly different from base case.
4-22
IllUbii
-------
1972
1973
1974
Scale
one inch = 12.
Figure 4-12.
First Quarter Wind Roses
Logan International Airport
4-23
Waklen
-------
As noted, the average increase in concentration due
to the actual variance component was 2.7 percent, and the increase in con-
centration due to the hypothetical full variance component was 52 percent.
Table 4-4 presents the relative contribution of three source categories
(power plants, large* residual oil users other than power plants, small
residual oil users and all distillate users) to the increase in S02 con-
centrations observed during actual variance implementation and hypothetical
full variance implementation case above the no variance case. Data are
presented for four individual receptors positioned to represent spatial
variation and for two receptors positioned to indicate the effect in the
vicinity of two large power plants which were granted variances.
These data indicate that distillate oil and small
residual oil users contributed about 75 percent of the increase in con-
centration due to actual variance implementation; power plants contributed
about 15 percent. With full variance implementation, the small users
would contribute about 60 percent and power plants 30 percent.
(2) Total Suspended Particulates
The principal sources of particulates in the Metro-
politan Boston region are incinerators, mobile sources and combustion
sources. Neither incinerators nor mobile sources were significantly af-
fected by conservation efforts during the energy shortage, nor are their
emission rates a function of heating requirements.** The only variance
granted which significantly affected particulate emission rates among fuel
combustion sources in the Metropolitan Boston AQCR was the allowance for
burning coal in three units of New England Power Company's Salem Harbor
power plant.
Table 4-5 presents the average percentage change in
TSP concentrations between the first quarter of 1973 and the first quar-
ter of 1974 based on model predictions for the base grid (Simulations (1)
and (3) described in Section IV.B.3).
*
>_ 20 million gallons/year.
**
i.e., they do not vary significantly with heating degree-days.
4-24
Maiden
-------
TABLE 4-4
COMPONENTS OF CHANGE IN S02 CONCENTRATIONS
RESULTING FROM VARIANCES BY SOURCE CATEGORY
Actual Implementation
Hypothetical Full
Variance Implementation
Region of AQCR
West
Central Business
District
North
Coastal
Point Source Dominated
3 km ESE of Edgar
Plant
3 km ESE of Salem
Harbor Plant
Net
Change
(Percent)
2.0
4.5
1.7
4.4
1.8
9.7
(Component
pp+
14.3
14.8
29.2
18.7
25.0
88.3
% of
LR*
3.6
8.5
0.0
8.7
2.5
0.8
Net Change)
SR+D**
82.1
76.7
70.8
72.6
72.5
10.9
Net
Change
(Percent)
40.8
62.1
44.3
97.2
73.1
74.0
(Component
pp+
32.2
37.3
29.3
58.3
55.0
45.7
% of
LR*
2.2
7.7
3.7
10.7
2.1
3.8
Net Change)
SR+D**
65.6
55.0
67.0
31.0
42.9
50.5
ro
en
Power plants
**
Large residual oil users other than power plants
Small residual oil users and all distillate oil
-------
TABLE 4-5
TSP Component
Net Change Growth and
1973-1974 Weather Conservation
Mean Percent Change -20.0 -20.3 -0.5
Standard
Deviation 20.5 19.4 1.3
Variances
Implemented
0.9
1.5
Also included in this table is a dissection of the total change into the three
components of interest: meteorology, growth and conservation, and variances
implemented. These results indicate that the average net change in TSP con-
centrations was a 20 percent decrease between the first quarter of 1973 and
the first quarter of 1974. This change results almost entirely from a 20
percent decrease in concentration due to differences in meteorology. Growth
and conservation and variances granted had a negligible impact on changes in
TSP concentrations during this period.
Table 4-6 compares these relative components with
those of two selected geographical areas - downtown Boston and the Salem
Harbor plant area. These data indicate that meteorological factors do not
influence the TSP concentrations in urban areas as strongly as they do for
the base grid system. Furthermore, comparison with similar data for S0~ in-
dicates a very small net change due to the weather component for this pollu-
tant. A qualitative analysis of these changes between 1973 and 1974 indicates
that the effects of weather were a composite of two opposing effects: the
tendency of the change in wind-stability patterns was to decrease concentra-
tions, and the effect of degree-days was to increase concentrations. S0£
concentrations, which are strongly tied to space heating requirements, showed
a balance between these factors. TSP concentrations, however, are affected
considerably less by change in degree-days and without this balancing factor
showed a significant drop in the total weather component. Urban TSP concen-
trations are more strongly influenced by space heating requirements due to the
concentration of dwelling and commercial establishments and thus are more nearly
balanced by these two components.
4-26
Maiden
-------
TABLE 4-6
PERCENT CHANGE IN TSP CONCENTRATIONS
1st QUARTER 1973 TO 1st QUARTER 1974
BASE RECEPTOR GRID
Components
Receptor
Grouping
Base
Urban Locations
Salem Harbor Area
Net Change
1973-1974
-21.0
-11.6
14.6*
Weather
-20.3
-10.5*
8.7*
Growth and
Conservation
-0.6
-1.1
3.9*
Variances
Imple.
0.9'
0.0
1.9
Full Variance
Implementation
0.9
0.0
1.9
* Significantly different from base case.
4-27
Ulaiden
-------
Meteorological effects in the vicinity of the Salem
Harbor plant produced an increase in TSP concentrations similar to the in-
crease in S02 observed at coastal stations. The variance to allow coal burn-
ing at this plant had little effect on TSP concentrations due to the effici-
ency of its electrostatic precipitators and its coastal location.
Full variance implementation would not have signi-
ficantly changed TSP concentrations.
4-28
/Ukkfa
-------
V. REFERENCES
1. "EPA's Position on the Energy Crisis", Environmental News, EPA, Wash-
ington, D.C., January 1974.
2. Morgenstern, P., et al, Air Pollutant Emission Inventory for the
Metropolitan Boston APCD, Maiden Research Corporation, Cambridge, Ma.,
June 1972.
3. Federal Register, Volume 37, p. 10842.
4. Bendersky, M.S., "Air Pollution Modeling's Role in the Change of
Massachusetts Bureau of Air Quality Control Regulation 5.1.3", Mas-
ter's Thesis, Massachusetts Institute of Technology, Cambridge, Ma.,
October 1974.
5. Federal Register, Volume 38, p. 34116.
6. Federal Register, Volume 39, p. 3822.
7. Federal Register, Volume 39, p. 17441.
8. Federal Register, Volume 39, p. 15272.
9. Federal Register, Volume 39, p. 32807.
10. Joly, G.T., Air Pollution and the Energy Crisis in Massachusetts,
presented at the New England Air Pollution Control Association Meet-
ing, May 25, 1974, Boston, Mass.
11. Transportation Controls to Reduce Motor Vehicle Emissions in Boston,
Massachusetts. Office of Air Programs Publication No. APTD-1442,
Environmental Protection Agency, Research Triangle Park, N.C., December
1972.
12. Shiskin, J., Young, A.H., and Musgrave, J.C., The X-11 Variant of
the Census Method II Seasonal Adjustment Program. Technical Paper No.
15, Bureau of the Census, U.S. Department of Commerce, Washington,
D.C., February 1967.
13. Spirtas, R., and Levin, H.J., "Patterns and Trends in Levels of Sus-
pended Particulate Matter", APCA Journal, Volume 21, p. 329, 1971.
14. McLaughlin, R.L., Time Series Forecasting, Marketing Research Tech-
nique Series No. 6, American Marketing Association, 1962.
15. Brown, R.G., Smoothing, Forecasting and Prediction of Discrete Time
Series, Prentice Hall, Inc., Englewood Cliffs, N.J., 1963.
5-1
Maiden
-------
16. User's Manual; SAROAD. Office of Air Programs Publication No. APTD-
0663, Environmental Protection Agency, Research Triangle Park, N.C.,
November 1971.
17. Snedecor, G.W., and Cochran, W.G., Statistical Methods (6th_ Edition),
Iowa State University, Ames, Iowa, 1967.
18. Dixon, W.J., and Massey, F.J., Introduction to Statistical Analysis
(3rd. Edition), McGraw-Hill, New York, 1969.
19. Draper, N.R., and Smith, H., Applied Regression Analysis. Wiley, New
York.
20. Miller, R.G., Simultaneous Statistical Inference. McGraw-Hill, New
York, 1966.
21. Box, G.E., and Jenkins, G.M., Time Series Analysis Forecasting and
Control; Holden Day, San Francisco, Ca., 1970.
22. Benesh, F.H., and Siege!, R.D., Development of the 1972 Area Source
Inventory for the Commonwealth of Massachusetts. Prepared for the
Commonwealth of Massachusetts and the Environmental Protection Agency,
Contract No. 68-02-1373, Task Order 3, Cambridge, Mass. (April 1975).
23. Sales of Fuel Oil and Kerosene in 1970, 1971, 1972, and 1973, Mineral
Industry Surveys, Bureau of Mines, U.S. Dept. of Interior, Washington,
D.C.
24. Compilation of Air Pollutant Emission Factors, Environmental Protection
Agency, Office of Air Programs, Publication No. AP-42, Research Triangle
Park, N.C., April 1973.
25. An Evaluation of Control Strategies for Stationary Fuel Burning Sources
the Thirty Inner Cities and Towns of the Metropolitan Boston Intrastate
Air Quality Control Region. Prepared by Wai den Research Corp. for the
Bureau of Air Quality Control, Commonwealth of Massachusetts and the
Environmental Protection Agency, Office of Air Programs, Raleigh, N.C.
(June 1973).
26. Commonwealth of Massachusetts, Dept. of Environmental Affairs, Population
and Employment Projections 1970-1985, Boston, Mass., 1975.
27. Benesh, F.H., and Chng, K.M., Methodology Development and Data Collection
to Update the NEDS Area Source Bank, in preparation for Environmental
Protection Agency, Contract No. 68-02-1410, Cambridge, Mass., 1975.
28. Air Quality Display Model, U.S. Dept. of HEW, PHS, NAPCA, prepared under
Contract No. PH 22-68-60 (November 1969).
5-2
UJalden
-------
29. Air Quality Implementation Planning Program, U.S. Dept. of HEW, PHS,
NAPCA, prepared under Contract No. PH 22-68-60.
30. Martin, D.O., and Tikvart, J.A., "A General Atmospheric Diffusion
Model for Estimating the Effects on Air Quality of One or More Sources",
APCA Journal (June 1968), pp. 68-148.
31. Environmental Protection Agency, Guide to Compiling a Comprehensive
Emissions Inventory, APTD-1135, Research Triangle Park, N.C.
32. Yankee Oilman, Vol. 20, #6, New England Fuel Institute, Boston, Mass.,
October 1974.
33. Yankee Oilman, Vol. 19, #11, New England Fuel Institute, Boston, Mass.,
March 1974.
34. "Energy-Saving Success or Flop? US Isn't Sure," Boston Globe, Boston,
Mass., February 16, 1975.
35. Commonwealth of Massachusetts, "Review of Alternative Strategies for the
Attainment of the Primary and Secondary Ambient Air Quality Standards
for Sulfur Diox4de in the Metropolitan Boston AQCR," Draft Report, BAQC,
Boston, Mass., 1975.
5-3
Maiden
-------
VI. APPENDICES
A. AIR QUALITY ANALYSIS
1. Data Analysis
a. X-ll Ratio to Moving Average Analysis
(1) Methodology
Air monitoring data, for ambient concentrations of
S02 (13 sites) and TSP (7 sites) in the Metropolitan Boston AQCR, was
judged sufficient for time series analysis (see Table A-l). Time series
of monthly pollutant concentrations measured at these sites were decom-
posed into three components or movements using a ratio to moving average
technique. These movements are: a trend-cycle component T which is com-
prised of both long-term and short-term trends, a seasonal com-
ponent S that quantifies the persistent periodic behavior in the series
that can be related to the temporal changes from one month to another,
and an irregular component I which quantifies the nonperiodic and
unpredictable part of the series. The X-ll Variant of the Census Method
II Seasonal Adjustment Program [12] was used to perform this analysis.
This program is based on a complex ratio to moving average technique.
Note that an a priori concept of what the "trend" of the data should be
was not assumed in this analysis. Rather, the trend-cycle component was
defined by the analytical procedure.
The analysis was performed on the original air quality
data using both a multiplicative (T*S*I) and an additive (T+S+I) model of
the series components in order to assess the most appropriate model in
each case. The decision criteria used in this assessment was to deter-
mine which model produced the smallest irregular component in each case.
To measure the magnitude of the irregular component, the final irregular
series standard derivation expressed as a percentage of the original series
mean value was calculated.
Ulalden
-------
TABLE A-l
METROPOLITAN BOSTON AIR QUALITY MONITORING SITES
RECORDING SUFFICIENT DATA FOR TIME SERIES ANALYSIS
IN THE INDICATED TIME PERIODS
SAROAD*
Site Code
Site
Location
Description
Series
Name
Tirre Span of Data
1970-74 1971-74
Urban
Core
Sites
sp_z
0240001(F01)
0240002(F01)
0240012
0240013
0340001
1160001
1200001
1220002
1480002
1700001
1880001
2340003
2620002
JSP
0240002(F01)
0340001
1200001
1480002
1700001
1880001
2340003
Government Center, Boston
Kenmore Square, Boston
South Bay, Boston
Central Square, East Boston
Greenough St., Brookline
Village St., Marblehead
US Army Site, Maynard
Main St., Medford
Dedham Ave., Needham
Nahatan St., Norwood
Hancock St., Quincy
Beaver St., Waltham
Montvale Ave., Woburn
Kenmore Square, Boston
Greenough St., Brookline
US Army Site, Maynard
Dedham Ave., Needham
Nahatan St., Norwood
Hancock St., Quincy
Beaver St., Waltham
Center City-Commercial S02A
Center City-Commercial S02B
Suburban-Industrial S02C
Center City-Commercial S02D
Center City-Residential S02E
Center City-Commercial S02F
Rural-Agricultural S02G
Center City-Commercial S02H
Center City-Residential S02I
Center City-Residential S02J
Center City-Commercial S02K
Rural Agricultural S02L
Center City-Commercial S02M
Center City-Commercial TSP1
Center City-Residential TSP2
Rural-Agricultural TSP3
Center City-Residential TSP4
Center City-Residential TSP5
Center City-Commercial TSP6
Rural-Agricultural TSP7
v/
i
Storage and Retrieval of Aerometric Data
-------
The X-ll program produced three potential trend in-
dicators — the final seasonally adjusted series (T*I or T+I), the months
for cyclical dominance series (MCD), and the final trend-cycle series (T).
These are labeled (1), (2), and (3), respectively, in Figure A-l which
compares the characteristics of these three trend indicators for a typ-
ical economic time series [14]. Referring to Figure A-l, it is important
to note that the fluctuations in series (1) are dominated by the
irregular component, whereas those in series (3) are dominated by the
trend component. In (3) a longer time span is used in computing the mov-
ing average than in (2) or (1) and as the time span increases, the trend
component grows more important relative to the irregular and eventually
dominates it. The crossover point is called the "months for cyclical dom-
inance", and the span associated with this point is used to derive the MCD
curve (2). Obviously, the smoothest of the three trend indicators, the
final trend series (3) is excellent for historical purposes. By contrast,
the seasonally adjusted series is the best of the series for analyzing
very recent trends in data because of its sensitivity to short span fluc-
tuations in the data.
(2) Time Series Component Graphs
Significant changes were made to the printed output
from the Census Bureau X-ll ratio to moving average analysis program to
provide a complete set of graphs for all series components of interest.
As an example, series component graphs from the multiplicative model for
the six representative data series (see Table A-2 and section II.A.2.a)
are shown in Figures A-2 through A-19. An index to the content of these
graphs is listed in Table A-3. All concentrations shown are in yg/m3.
Note that although graphs are shown for only the results from the multi-
plicative model for the six series, the X-ll ratio to moving average an-
alysis was undertaken using both the additive and multiplicative models
applied to all twenty original series (see Table A-l).
UlaJden
-------
Figure A-l. A Comparison of the Characteristics of the
Trend Series Indicators for a Typical Economic
Time Series [14].
1946 1947 1948 1949 1950 1951 1952
1953 1954 1955 1956
Time
1957 1958 1959 1960 1961
-------
TABLE A-2
REPRESENTATIVE METROPOLITAN BOSTON AIR QUALITY MONITORING SITES
CHOSEN FOR DATA ANALYSIS
CD
SAROAD*
SHe Code
Location
Site
Description
Series
Name
Time Span
of Data
S02
0240002(F01)
1700001
2340003
JSP
0240002(F01)
1700001
2340003
Kenmore Square, Boston
Nahatan St., Norwood
Beaver St., Waltham
Kenmore Square, Boston
Nahatan St., Norwood
Beaver St., Waltham-
Center City-Commercial S02B
Center City-Residential S02J
Rural-Agri cultural S02L
Center City-Commercial TSP1
Center City-Residential TSP5
Rural-Agricultural TSP7
January 1970-May 1974
January 1971-June 1974
March 1971-June 1974
January 1970-June 1974
January 1971-Ju.ne 1974
March 1971-June 1974
Storage And Retrieval of AerometHc Data
-------
Figure A-2
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES S02B
-------
a
o
S
i—i
sf
ai
— o
Of
<
-5
;IT\
O
LJ
O •-
o- •
o
•*•!
O
Ol
•a
Oi X
•»•! • ~
00
IN); •
•O
i/V
»v
-f
,si
U. UJ
O -I
o
-Ul>-
I oio
: aiio
- 3! i
•^ M-
o
2
oi
-> _i
''r
N -•
X O
» i^
<
O
00
Figure A-3
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES S02J
• o
o
^r
'o
• co
CM
!0
• i-O
>|3
I •
!»
• IINJ
-------
Figure A-4
o •
o
• o
o
m
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES S02L
o !
-------
Figure A-5
oo • X'
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES TSP1
-------
o
o
(M
-------
o
m
-------
I I
figure A-8
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES S02B
-------
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES S02J
-------
o •
o
o
Figure A-10
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES S02L
• o
o
o
o
> o
o
t'O
!O
•O
o
i 03
O
• OJ
-------
CO
* •
CM
Figure A-ll
I
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES TSP1
-------
i !
Figure A-12
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES TSP5
. *
-o
A-16
. ;eo
KM
-------
o
5f
Figure A-13
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES TSP7
I-
o
n
I I
-------
I '
~ "
(
_J J1 J
U. ILL
! I
Figure A-14
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES S02B
CO X
O
O O O'O
r—jr— r~jr» h-
Q. I- > O Z
LU O O|LU <
i^'D ? O -,
-------
sL
-
I-W
-""I
o —I
a: O
N HO i
-" Jl O !
3 01 -I !
15
"T
-~ —a 01
' a. < o 3
:i < s: -5 -j
Figure A-15
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES S02J
tO
O
,13}
.IS
c
-T
-------
I I
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES S02L
. o
o
-------
Figure A-17
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES TSP1
-------
o
in
IN
o
o
Ujl
— I
DC1
o
in
OJ
-J l/>
iO «-
-}
3 a.
-^
XI U.IOJ
<* Oi-J
I- •->
JU >•
O'O
< z
" JJ'LU I
~" 9 z
_>
-M -.1 J3
Figure A-18
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES TSP5
rf X
< —I
I
O
I I -•
5T
—. -« UJ
» l^
uJ in
I CM
o
o
o
in
CD orr itx
UJ < o.
LL s:
•*•!•*•
r- ir-
>- z
-------
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES TSP7
-------
TABLE A-3
INDEX TO SERIES COMPONENT GRAPHS PRODUCED BY THE
X-ll RATIO TO MOVING AVERAGE ANALYSIS (MULTIPLICATIVE MODEL)
FOR SIX REPRESENTATIVE TIME SERIES
Series Components Graphed Figure Number
Original Series (0) & Trend-Cycle Component (T) A-2 - A-7
Seasonal Component (S) & Irregular Component (I) A-8 - A-13
Seasonally-Adjusted Series (T*I) & Trend-Cycle A-14 - A-19
Component (T)
A-24
-------
(3) Results
The first task in evaluating the results of the X-ll
analysis was to determine which model produced the smallest irregular com-
ponent in each case. To measure the magnitude of the irregular component,
the final irregular series standard deviation as a percentage of the original
series mean value was calculated. The results, shown in Table A-4, reveal an
additive model to be most appropriate to the S02 data 69% of the time (9 of
13 cases), while the TSP data were modeled best by a multiplicative model 86%
of the time (6 of 7 cases). It is interesting to note that the S02 data are
arithmetic monthly averages of daily measurements, while the TSP data are geo-
metric monthly averages. The difference between the two data reduction pro-
cedures is due to the manner in which Federal and Massachusetts Ambient Air
Quality Standards have been specified. Although one model performs slightly
better than the other in each instance, the trend-cycle and seasonal component
results are comparable in each case. Therefore, in analyses to follow, the
additive and multiplicative forms were used interchangeably. In some instances,
it will be seen that only the multiplicative model is used. Since the seasonal
and irregular components are temporal functions with a mean value of 1.0, the
magnitude of the multiplicative trend-cycle component is directly comparable
to that of the original time series.
Next, the X-ll program results were evaluated to determine
whether any of the data were not amenable to statistical trend analysis. In
this regard, the X-ll program applied an analysis-of-variance F-test for stable
seasonality to determine whether a true seasonal component was present in the
data series. The results of this test distributed by model configuration are
shown in Table A-5. One cause of lack of stable seasonality in measured data
can be that monthly averages are based on too few measurements. For the ana-
lyzed data, the individual data points that were used in the computation of the
monthly averages were scanned to ascertain if any time series should be elimi-
nated from further analysis due to an inadequate amount of measurements or due
to dominance by the irregular component. The results indicate that none of the
data series are based on too few measurements or dominated by an irregular com-
ponent to the extent that they are totally unreliable. However, it should
be noted that, at most monitoring sites, only a small number of daily mea-
surements were available for use in computing the monthly average. Except
A-25
UJalden
-------
TABLE A-4
FINAL IRREGULAR SERIES COMPONENT STANDARD DEVIATION
AS A PERCENTAGE OF THE ORIGINAL SERIES MEAN VALUE
Series Name
S02A
B
C
D
E
F
G
H
I
J
K
L
M
TSP1
2
3
4
5
6
7
Model
Multiplicative
64.0
30.8
40.8
43.5
50.1
83.0
141.2
62.8
69.4
30.2
62.5
84.8
51.6
10.1
42.3
21.6
15.1
27.9
15.0
13.6
Additive
43.9
23.1
40.2
50.3
70.4
51.9
53.0
43.9
49.2
41.9
44.9
73.3
87.7
9.9
70.5
29.3
15.3
46.4
16.4
14.1
Recommended
Model
A
A
A
M
M
A-
A
A
A
M
A
A
M
A
M
M
M
M
M
M
A-26
-------
TABLE A-5
TIME SERIES FOR WHICH STABLE SEASONALITY WAS PRESENT
AT THE 1% SIGNIFICANCE LEVEL
USING THE INDICATED MODEL CONFIGURATION
Series Name Multiplicative Additive
|M A .
S02A / /
B / /
C / /
D /
E /
F
G
H
I / /
0 / /
K / /
L
M
TSP1 / /
2
3 / /
4 / /
5
6 / /
7 / /
A-27
Walden
-------
at the Kenmore Square site, the number of daily measurements per month ranged
from 3 to 8 in 1970 - 1974. By contrast, an average of 19 measurements
formed the basis for Kenmore Square monthly concentrations. Because of
this larger data base, the time series data from Kenmore Square contains
a smaller irregular component than that from other sites (see Table A-4).
Evaluation of the long-term trend in a sequence of
air quality parameters, such as annual average concentrations is impor-
tant in order to assess the effects of relevant causal factors. Tables
A-6 and A-7 show, respectively, the annual averages of ambient S02 and
TSP concentrations measured at the monitoring stations in the AQCR from
which data was available. The averages are also given for the series
trend-cycle components derived using a multiplicative model of the X-ll
ratio to moving average technique. Figures 2-4 and 2-5 show composite
annual average S02 and TSP trend-cycle components from the multiplicative
model in the AQCR; data from sites in the innermost core area of Boston
are graphed separately. Trend-cycle components for the data series re-
vealed that, in general, regional S02 levels fell consistently during
the entire time period of analysis, with the steepest decline occurring
prior to 1972 (see Figure 2-4). Ambient TSP levels also exhibited a gen-
eral downward trend similar to that of S02 levels, but not as pronounced
or consistent (see Figure 2-5).
The hypothesis that long-term changes in ambient S02
and TSP levels are caused principally by changes in annual meteorological
conditions, and not by fuel use regulations or other factors, implies a
strong positive correlation should be found between annual average con-
centrations and total heating season degree-days. The annual degree-day
totals as measured at Boston's Logan International Airport are shown below:
Year Total Degree-Days
1970 5,852
1971 5,738
1972 5,912
1973 5,139
A-28
Ulatien
-------
TABLE A-6
ANNUAL AVERAGE S02 GONCFNTRATIONS MEASURED
IN THE METROPOLITAN BOSTON AQCR (ug/m3)
Data
Series 1970
S02Af 86
SO^B1" 167
S02Cf 95
S02Df 90
S02E
S02F
S02G
S02H
so2i
S02J
S02K
S09L
L
S02M
Original Series
1971
50
89
51
61
47
22
11
42
19
27
36
19
23
1972
24
68
34
34
23
16
8
21
10
17
36
11
13
1973
29
32
26
14
23
13
7
18
10
15
26
10
12
1974* 1970
23 108
26 170
16 101
12 82
15
13
5
22
12
22
23
17
18
Trend-Cycle Component**
1971
53
82
51
54
45
20
14
40
23
32
38
22
30
1972
17
71
32
36
25
15
6
23
9
17
28
9
12
1973
29
29
27
14
19
11
7
20
9
17
30
8
12
1974*
22
23
13
9
15
15
5
18
9
16
18
15
16
**
Does not contain data beyond June 1974.
From a multiplicative ratio to moving average model.
Urban core site.
A-29
Waiden
-------
TABLE A-7
ANNUAL AVERAGE TSP CONCENTRATIONS MEASURED
IN THE METROPOLITAN BOSTON AQCR (yg/m3)
Data
Series
TSP1 f
TSP2
TSP3
TSP4
TSP5
TSP6
TSP7
Original Series
1970 1971 1972 1973
158 114
87
33
49
50
66
56
113
47
24
39
44
54
41
104
46
36
38
61
49
55
1974*
89
36
22
31
43
48
32
Trend-Cycle Component**
1970 1971 1972 1973
151 117
90
33
48
57
66
52
112
46
25
39
46
55
42
102
46
31
37
49
50
55
1974*
80
36
21
28
40
46
33
Does not contain data beyond June 1974.
**
From a multiplicative ratio to moving average model.
Urban core site.
A-30
lUUM
-------
Practically no correlation exists between annual S02 and TSP levels and
degree-day totals. A prediction based solely on degree-day data would
forecast the peak annual concentrations of S02 and TSP to occur in 1972,
a fact which is not borne out by the measured data. Consequently, the
observed long-term changes in ambient S02 and TSP levels cannot be at-
tributed to corresponding changes in this meteorological parameter.
b. Exponential Smoothing —Confirmation of X-ll Trend
Components
(1) Methodology
The results of the ratio to moving average analysis
were exploratory, allowing identification of trends and changes in trends
in the measured data. The next step in the data analysis was to confirm
the results of the moving average analysis in identifying and quantifying
trends. The technique selected to supply this confirmation was triple
exponential smoothing [15]. The smoothing technique was applied to the
seasonally adjusted series. Alternatively, the original series could have
been used, but then a separate exponential smoothing step just to remove
seasonality would have been necessary.
Exponential smoothing is an analytical technique for
estimating trends by weighting all previous observations. By this tech-
nique, weighting factors decrease exponentially, being largest for obser-
vations in the most recent past. Exponential smoothing produces an average
time series in which past observations are geometrically discounted ac-
cording to age. By contrast, an N-month moving average weights the N most
recent observations each 1/N, and all earlier observations have weight
zero. These two basic smoothing techniques are compared in Figure A-20
which shows the weights used to obtain a 12-month moving average time
series (equal) and a triple exponentially smoothed time series (exponen-
tial). Note the two examples shown do not necessarily produce equivalent
results. The exponential smoothing weights shown in Figure A-20 were
derived using a smoothing constant of 0.10.
A-31
UJalden
-------
NO 319-C Mill MITJR-. 1 -10 LM 2 Do UIVISIONS.
FIGURE A-20
IN bTOCK DIRECT FRO'
. .-K co . NO^woon MASS O2O62
Two Ways of Weighting Data in Smoothing Discrete Time Series
N is the reciprocal of the weighting factor
used in a moving average
otis the smoothing constant used in
exponential smoothing
Exponential Weights,OC= 0.10:
Equal Weights, N=12
-r-H-7--h-
-------
Selection of a suitable smoothing constant is an im-
portant part of any trend analysis employing exponential smoothing as this
parameter controls the response sensitivity of the process. The smoothing
constant is defined in the range 0.0 to 1.0. The lower the value of the
smoothing constant, the higher the degree of data smoothing obtained. The
rate of response (i.e., sensitivity) to a changing trend improves with a
higher smoothing constant, but this virtue is mitigated by a decreasing
ability to smooth out random fluctuations. These two properties must be
balanced against one another in choosing an appropriate smoothing con-
stant. In the current application, the value that was found to produce
adequate smoothing of the data and sufficient response sensitivity by
visual inspection varied between 0.10 and 0.25, depending on the degree
of irregularity of the data. It should be noted that though exponential
smoothing is analytically a more sensitive technique than moving averages,
a slight lag in the smoothed data is always produced as a result of smooth-
ing out random data fluctuations.
Triple exponential smoothing makes the assumption
that the process represented by the discrete time series (X), in this case
ambient S02 and TSP levels, can be modeled in time (t) using a quadratic
function:
X(t) = a + bt + ct2
The values of the model coefficients a, b, and c change stepwise throughout
the smoothing analysis and thus provide a predictive model based on the
most complete information at any given time step in the analysis. This
predictive ability was used in the statistical inference analysis to per-
mit testing of statistically significant changes in trends (i.e., corres-
ponding to changes in fuel regulations or implementation of variances).
Since exponential smoothing is a continuous process, it requires specifi-
cation of a set of initial model coefficients. There are several estab-
lished formulas for estimating initial values of a, b and c. None of
these starting techniques produced acceptable results in that the smoothed
series did not begin to follow the general directions of the raw data
A-33
Maiden
-------
until 10 or 20 time steps (months) into the analysis. The cause of these
failures was the large initial drop in pollutant levels in many of the
series. Therefore, a unique technique was devised whereby each time series
was smoothed in reverse initially, using one of the established starting
techniques to determine a set of final model coefficients. These coeffici-
ents were then used to forecast the series forward using negative values of
the time parameter, t. Alternatively, if the sign of b is reversed, the
coefficients provide acceptable initial values for the exponential smoothing
analysis.
The X-ll program used in the time series component
decomposition process produced three potential candidate series for the
exponential smoothing: the final seasonally adjusted series, the months
for cyclical dominance series, and the final trend series. Each of these
series are discussed in more detail in Section VI.A.l.a.(l). The final
seasonally adjusted series was judged the best series to use in confirming
the ratio to moving average trend analysis results through exponential
smoothing. The seasonally adjusted series is the original series with
only the seasonal component removed, whereas the other two series have had
portions of the irregular component removed as well. Since the seasonally
adjusted series contains the full irregular component, it provided the best
basis for a comparison of the trend lines produced by the exponential smooth-
ing and ratio to moving average analyses. As previously mentioned, the ori-
ginal series could have been used, but then a separate exponential smoothing
step just to remove seasonality would have been necessary.
The exact form of the quadratic exponential smoothing
function depended on whether the seasonally adjusted series to be smoothed
was derived in the X-ll program using an additive (T + I) or multiplicative
(T * I) model. In the current application, the following functional forms
were assumed:
X = (T + I) = a + bt + ct2 + E (Additive)
X = (T * I) = E * exp (a + bt + ct2) (Multiplicative)
where E is the error term.
A-34
UJalden
-------
(2) Results
Graphs of the triple exponentially smoothed seasonally
adjusted series and the final trend series for each of the six representa-
tive time series (see Table A-2) are shown in Figures A-21 through A-26
(additive model) and in Figures A-27 through A-32 (multiplicative model).
Although the results of the exponential smoothing analysis are shown for only
the six representative time series, this confirmatory analysis was undertaken
on all twenty original series (see Table A-l). On the basis of visual in-
spection, the exponential smoothing and X-ll ratio to moving average ana-
lyses produced comparable trend lines in all cases. These results justi-
fied use of exponential smoothing techniques to test for statistically sig-
nificant changes in air quality trends. This analysis was undertaken and
is described in Section VI.A.2.a.
Since exponential smoothing is a more sensitive ana-
lytical technique than the moving averages procedure, it is expected that
in cases where the irregular component dominated the analysis, the expo-
nentially smoothed series would exaggerate trend component minima and
maxima. Table A-8 gives a listing of the relative irregular component
magnitudes for the series shown in Figures A-21 through A-32. Such dis-
tortions did occur in the data in Figures A-22, A-23, and A-25 for the ad-
ditive model, corresponding to the series for which the irregular com-
ponent standard deviation as a percentage of the original series mean value
exceeded 40%. In Figures A-27 and A-32, such distortions are not readily
comparable, due to the non-linearities of the semi-logarithmic scale of
the graphs. As stated in Section VI.A.l.a.(3), the monthly averages were
sometimes based on just a few observations. In these instances, the ori-
ginal series possess large fluctuations, and it is hard to detect trends
in the data. The distortions observed in Figures A-22, A-23, and A-25 are
probably due to this sample size problem.
c. Seasonal Patterns in Ambient Particulate Levels
In exploring the observed trends in the data, comparisons
were made with the TSP data and results of an earlier study [13] of
nationwide TSP levels. That study analyzed ten years of data (1957-1966)
from the National Air Surveillance Network (NASN), also using a ratio to
A-35
Ulalden
-------
Figure A-21
i
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES S02B
BEST SMOOTHING CONSTANT = 0.20
o o o o o
r- r- r-
z m ee. ef. >
< OJ < Q. <
r- f» t*- r~ !r- r-
r- t>-
z -i o a. t- >
3D UJ U O
-» -> < (^ o Z
2 «- x
z _i o a-1-
3 3 3 UJ O
oc ce. >
Ul < Q. <
o -> u. r •
-------
LJ
O
2
>
IU
t-il
a
iu
2
5'
2
o
iu
I U M
iti O
It N M
ee p« o
K O O
'!> i?
o •
f «O
fcx
f-
U
(9
O *
c
u o
r-ir- f-
z!eo ee
'
a: >•
ft. <
Figure A-22
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES S02J
BEST SHOOTHING CONSTANT * 0.20
• -t
.IS
IN
. (M
! . I
-------
Figure A-23
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES S02L
BEST SMOOTHING CONSTANT = 0.25
o O-'H > o r
5 UJ|0 O Lb
< (SI
-------
I
Figure A-24
i
It)
Q
UJ
>
g
o
UJ
z
8
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENHORE SQUARE, BOSTON SITE
SERIES TSP1
Oi
r-4
fM.
J
10 UJ
_J 10
>• 0 K
-*> z
JO HH
< p
IH O SL
Z UJ U.
UJ OC 0
Z K
O uj
ft. -I U
X < Z
uj Z uj
11 Q
Uj U. 1-4
-J 0
Q. . 2
"i rj 1-1
« ^" O
K Q (J
1 t l
O
VH
•
O
o>
0 •
•" >t
t- r-
UJ
X
X
^xo *
io
'K
iw
BEST SMOOTHING CONSTANT
O O O O'O O
r-
o o.»- >
S iuu o
-j •« v» o r
UJ 00
_i in
-------
Figure A-25
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES TSP5
c, •
O
o
a:
111
o
<
UJ
l/>
o
UJ
X
o
o
> o
J V
_J U
HO
• or o
O
a. -J
X <
UJ Z
•-4
UJ U.
_l
Q. •
IM (SI
IK
I"
x o
r
lo
0 0 •
2 1-1 (M
1-1 K CO
O UJ
U SE
I K
t-H
— tf.
* <
~ I •
UJ IT
U
c/i
BEST SMOOTHING CONSTANT = 0.20
M
• c
-------
Figure A-26
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES TSP7
•Ol •
•Oi •
in! • O
I/)
UJ
a.
UJ
i-i l/>
8 : <
Q uj
1 O
I UJ
1 X
m
o
D
• o >-
-J >• z
—I O t->
< o
1-1 Q Q.
t- z •
Z UJ U. _ (M
tf *H O UJ
1 K Q o a:
i . ifE
iac
< <
X
<-)
IN
UJ
-------
o •
o
ID
Figtire A-27
o
-o «
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES S02B
SMICTHINii CONSTANT =
-------
Figure A-28
I
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES S02J
bEST SNCOTHING CONSTANT
in
• ro
•
O
-------
Figure A-29
o
rt •>• Z _i O 0.
Q. < r> o ^ uj
-5 <*
(M IM rvl
N- r- N- h-'p- r-
t- > o z a) of
o o LU u- n
S02 AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES S02L
SKOTHING CONSTANT = 0.20
f\j rsi fxj fsjirvj (N
r~ r- r>- r^ir- r-
ce. >-(Z _i:^ Q.
Q. <••=> 3'3 "J
< x -n -» < oo
•«*•
• (M
'CO
-------
Figure A-30
r\j
r- •
IN
00
rfi
IN
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES TSP1
dLM SfCOTHING CONSTANT = 0.20
-------
Figure A-31
•o
• (*>
1
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES TSP5
BEST S^COTHING CONSTANT
'(M
• O
-------
2!
UJ|
CO
Q
UJ
CO t
• o *-
-1
-J
<
-r
JJ
Z
'3
0.
X
OJ
XI
Q.
••*
tt
•-
1
an x
r ~
u
4
s
o
1
>• Z
•J -«
•3i
O Q.
ULL J
2C ^ O
>- V
JJ O
-1 -)
Z JJ J
J. -i X
« z;
N « J>
0 0,_|
1
1 1 —
— > -~,LU
0 » 01
LU
O
c/1
1
I .
Figure A-32
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES TSP7
SNCOTHING CONSTANT = 0.25
(M
• r-
to
-------
TABLE A-8
FINAL IRREGULAR SERIES COMPONENT STANDARD DEVIATION
AS A PERCENTAGE OF THE ORIGINAL SERIES MEAN VALUE
Series
Name
Irregular Component
(Percentage)
Figure Number of Triple
Exponentially Smoothed Series
Additive Model
S02B
S02J
S02L
TSP1
TSP5
TSP7
Multiplicative Model
S02B
S02J
S02L
TSP1
TSP5
TSP7
23.1
41.9
73.3
9.9
46.4
14.1
30.8
30.2
84.8
10.1
27.9
13.6
A-21
A-22
A-23
A-24
A-25
A-26
A-27
A-28
A-29
A-30
A-31
A-32
A-48
/UlaUen
-------
moving average technique. The NASN data revealed that ambient TSP con-
centrations at urban sites demonstrated seasonal patterns generally
characterized by high winter values and relatively lower summer values.
By contrast, seasonal patterns of non-urban sites were typified by high
summer peaks and low winter values.
The first column of Table A-9 gives a preliminary break-
down by SAROAD* site description of data used in the current study. These
categories are based on two descriptors [16]. The first descriptor de-
fines a sampling site as being either in a Center City, Suburban or Rural
area. For the second site descriptor it was necessary to estimate the
dominating influence within a 1-mile radius of the sampling site. For
Center City of Suburban, the following categories were possible:
(1) Industrial: implies product-oriented estab-
lishments such as manufacturing concerns,
utilities, mining, and graineries.
(2) Commercial: implies service-oriented estab-
lishments. A unique traffic pattern into and
out of the area would be expected. Retail
establishments, shopping centers, gas stations,
laundramats, etc., comprise this category.
(3) Residential: because many other areas are also
used residentially, this category is selected
only in the absence of a dominating industrial
or commercial influence.
For rural sampling sites, two other categories were used:
(4) Near Urban: category for samplers placed in a
rural area, yet close enough to a major urban
center to be materially affected by the urban
area.
(5) Agricultural: category encompassing orchards,
crop raising, cattle and sheep grazing, etc.
*
Storage and Retrieval of Aerometric Data.
A-49
UJaiden
-------
Table A-9
A Comparison of Actual Site Description and That Implied
by the Characteristics of the Seasonal Component of TSP Time Series
Data Collected in the Metropolitan Boston Region
Series
Name
TSP1
TSP2
TSP3
TSP4
TSP5
TSP6
TSP7
Consolidated SAROAD
Site Description
Urban
Suburban
Nonurban
Suburban
Suburban
Urban
Nonurban
Site Description Implied by Seasonal
Component Characteristics Using
Definitions from an NASN Study [13]
Urban
Nonurban
Nonurban
Nonurban
Urban
Urban
Nonurban
A-50
-------
It was desirable to consolidate these site descriptions into urban and
non-urban categories for purposes of comparison with the NASN results.
The technique used to accomplish this was to label Center City-Commer-
cial sites as urban, Rural-Agricultural sites as non-urban, and all other
sites (e.g., Center City-Residential) as suburban.
Figures A-33 through A-39 show the seasonal components (S)
of time series TSP1 through TSP7 using a multiplicative ratio to moving
average model. By comparing the time of year of the seasonal component's
peak with the NASN definitions, the second column in Table A-9 was con-
structed. Note that series TSP2 through TSP6 (Figures A-34 through A-38)
are characterized by double peaks, one in summer and one in winter. In
these cases, the site description listed as implied by the seasonal com-
ponent characteristics is the one that predominated. Based on the above
results, it is concluded that the observed seasonal patterns in TSP levels
confirm those found in the NASN study [13] of nationwide TSP levels.
d. Analysis of the Correlation Between the Seasonal
Component Produced by the X-ll Analysis and Meteo-
rological Variables
In order to ascertain whether significant interdependence
existed between measured heating degree-day or mean wind speed data and
the periodic components derived in the X-ll ratio to moving
average analysis of measured air quality data, linear regressions between
these variables were performed for the six representative time series
(Table A-2). The results are shown in Table A-10. It can be seen from
the correlation coefficients that significant interdependence does exist
in all series, except TSP5. The implication of this finding is that the
meteorological variables, degree-days and mean wind speed can be used as
a measure of the seasonality in ambient S02 and TSP levels.
Other important characteristics worth noting from Table
A-10 are that all the S02 series have a seasonal component positively
correlated with degree-days, i.e., both S02 levels and degree-days ex-
hibited a periodicity with maxima in the winter and minima in the summer.
A-51
Utalden
-------
FIGURE A-33
t
00
•a- «
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
KENMORE SQUARE, BOSTON SITE
SERIES TSP1
O
g
H
E-i
At
M
EH
J
D
S
-------
FIGURE A-34
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
GREENOUGH STREET, BROOKLINE SITE
SERIES TSP2
o
• t-
Pi
H
-------
FIGURE A-35
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
U.S. ARMY SITE, MAYNARD
SERIES TSP3
Q
i
M
H
s
M
"0 • U
_) LLJ
-•* "
X —
-J < Z
I -)
• ' • »
I 11 I —
I 3;
H- —•! —. —• UJ
,v x jo * oo
— — I
X
o
OJ -O
-------
FIGURE A-36
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
DEDHAM AVENUE, NEEDHAM SITE
SERIES TSP4
CO
• o
in
• rn
(N
• O
CM
A-55
se •>- z j o a.
Q. •« 35 O —I UJ
O 2 32 a:
U
O "J <
-> U- S"
a.
-------
FIGURE A-37
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
NAHATAN STREET, NORWOOD SITE
SERIES TSP5
3
i
3
Ot
M
£->
to
•e
-------
FIGURE A-38
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
HANCOCK STREET, QUINCY SITE
SERIES TSP6
-o
f>
1M
• o
-------
FIGURE A-39
-o
in
TSP AMBIENT AIR QUALITY LEVELS IN BOSTON AQCR
BEAVER STREET, WALTHAM SITE
SERIES TSP7
-t
(M
(Nl
-t
• N
M
i
H
s
•f >- 2 _l O Q
•^ T -> -> LJ
< z: -5 — i < to
A-
f\J Csj rsj rO r^ .^
>- > O Z CC Q.
O -3 :jJ <• IJJ ~> Li-
i
• O ^ 3D Tf
O T UJ j- -J-
< T
-------
TABLE A-10
LINEAR REGRESSIONS BETWEEN THE SEASONAL COMPONENT OF MEASURED
AMBIENT S02 AND TSP LEVELS DERIVED IN THE X-ll
RATIO TO MOVING AVERAGE ANALYSIS AND METEOROLOGICAL VARIABLES
Heating Degree Days
Data Series
X-ll Multipli
S02B
S02J
S02L
TSP1
TSP5
TSP7
X-ll Additive
S02B
S02J
S02L
TSP1
TSP5
TSP7
R**
cative
0.95
0.84
0.84
0.83
0.20
-0.71
Model
0.96
0.85
0.86
0.83
0.43
-0.75
Slope
Model
0.11
0.10
0.15
0.050
0.012
-0.051
0.081
0.020
0.017
0.054
0.016
-0.026
Intercept
48
52
32
79
95
125
-38
-10
-8
-24
-8
12
Mean
R**
-0.71
-0.61
-0.54
-0.56
-0.07
0.56
-0.74
-0.62
-0.56
-0.56
-0.30
0.59
Wind Speed*
Slope
-1940
-1530
-1910
-734
-88
827
-1450
-310
-232
-862
-253
416
Intercept
275
238
271
167
109
27
131
28
21
79
23
-37
**
Evaluated as the reciprocal of mean wind speed in the regressions.
Correlation coefficient: an absolute value of 0.31 or larger
corresponds to a level of significance at most equal to the 5%
level, i.e., there is at least a 95% chance that the correlation
is significant.
A-59
Maiden
-------
The TSP data, however, show two seasonal patterns. The TSP1 series (Ken-
more Square, Boston) exhibited a positive correlation with degree-days
which is characteristic of seasonal patterns observed at urban sites in
a study of nationwide TSP levels [13]. By contrast, the seasonal com-
ponent at the TSP7 series (Beaver Street, Waltham) exhibited the converse
behavior, i.e., with maxima in the summer months and minima in the winter.
This pattern is characteristic of TSP levels measured at rural sites in
the same nationwide study. These results are comparable with those given
in Section VI.A.I.e. Further analysis of these interdependence relation-
ships is given in Section VI.A.2.b.(3).
2. -Statistical Inference
a. Exponential Smoothing — Statistical Trend Analysis
(1) Methodology
Since 1970, the Commonwealth of Massachusetts Bureau
of Air Quality Control (BAQC) has promulgated several regulations limit-
ing the sulfur and ash content of fuels burned in the Metropolitan Boston
AQCR (MBAQCR). These are summarized in Table A-ll. During the winter of
1973-74, when shortages of conforming fuel appeared imminent, many vari-
ances were granted on these regulations by the BAQC to large fuel users
in the MBAQCR during the period of December 1, 1973 to May 15, 1974. A
procedure employing the forecast potential of triple exponential smoothing
was devised to permit testing of the statistical significance of changes
in trends of measured air quality data corresponding to the onset of fuel
use regulations and variances.
The technique used to test the effect of regulations
and variances on ambient air quality levels was to compare the actual seasonally
adjusted values after the implementation date of a regulation or variance with
those values predicted by a triple exponential smoothing model. These
latter values were computed under the assumption of no changes in air qual-
ity trends and this hypothesis was tested against the observed series.
A-60
lUHU
-------
TABLE A-11
A SUMMARY OF REGULATIONS ADOPTED TO CONTROL THE SULFUR AND ASH CONTENT
OF FUELS BURNED IN THE METROPOLITAN BOSTON AIR QUALITY CONTROL REGION
Effective Date of Adoption
Regulations in Effect in Metropolitan Boston AQCR
July 1, 1970
October 1, 1970
October 1, 1971
June 1, 1972
Ban on all open burning
Ash content of fossil fuels limited to 9% by dry weight
Residual oil sulfur content limited to 2.2% and to 1.0% in core*
Distillate oil sulfur content limited to 0.3%
Coal sulfur content limited to 1.5% and to 0.75% in core*
Residual oil sulfur content limited to 1.0% and to 0.5% in core*
Coal sulfur content limited to 0.7% and to 0.37% in core*
Ban on residual oil use at facilities with rated boiler capacities
of 3 million Btu/hour or less
Ban on solid fuel use at hand-fired facilities with rated boiler
capacities in excess of 150 thousand Btu/hour
*The 13 core cities and towns are: Arlington, Belmont, Boston, Brookline, Cambridge, Chelsea, Everett,
Maiden, Medford, Newton, Somerville, Wai than, and Watertown
-------
Statistical tests of significance were applied to the residuals (actual minus
predicted values) obtained from the above procedure [17, 18], Conclusions
were then drawn as to whether a statistically significant change in the trends
of measured air quality data occurred coincident with the onset of regulations
and variances.
The predictive ability of the exponential smoothing model
depends on the value of a smoothing constant which controls the response sen-
si vity of the process. Values which enable a high degree of data smoothing
tend to produce the best long-range forecasts. In order to ensure the model
predicted the trend of the air quality data, the values of the smoothing con-
stant used in the current application were those employed in the trend analy-
sis confirmation. On the basis of visual inspection, the predictive ability
of the model under these circumstances was judged adequate for no more than a
period of ten months (or ten time steps).
The analysis of testing for statistically significant
changes in trends of measured air quality data after a certain time, t, was
undertaken using the seasonally adjusted form (i.e., seasonalycomponent removed)
of the six representative time series listed in Table A-2. The original series
were not used, since it was felt that seasonality in the data would confuse
statistical test results. The form of the exponential smoothing model used,
i.e., multiplicative or additive, depended on whether the seasonally adjusted
series was derived using a multiplicative of additive X-ll technique. Thus,
the actual values were from the seasonally adjusted form of the original series:
Actual = T + I or T * I
The predicted values were from a seasonally adjusted series predicted by the
exponential smoothing model assuming no change in the trend, T1, after time t:
Predicted = T1 + I or T1 * I
The residuals used in the statistical tests were the Actual minus the Predicted
values. An example showing how the Actual and Predicted series compared is
shown in Figure A-40 for the TSP1 series (additive model) and for t = December
1, 1973.
It should be noted that the seasonally adjusted series pro-
duced by the ratio to moving average technique may have still contained some
seasonal effects, although tests for residual seasonality were not performed,
A-62
/Ulakknl
-------
Value of Series Components
en
o
---J
o
00
O
O
O
ro
o
CO
o
Figure A-40,
Comparison of Actual and Predicted Seasonally Adjusted TSP1 Series
after December 1, 1973
-------
due to the inconclusive results of the statistical tests. Residual seasonality,
in the data that were analyzed, may have caused apparent changes in air quality
trends where none actually existed. Therefore, particular care was taken in in-
terpreting the results of this analysis. The stepwiee multiple linear regression
provided a more objective analysis of the effect of regulations and variances on
ambient air quality levels and was used to expand on the following results.
(2) Results
Nonparametric statistical tests of significance, the Sign
test and Runs test [18], were applied to the residuals obtained from the previ-
ously described procedure. Each time series listed in Table A-2 yielded resi-
duals (corresponding to the ten-month predictive ability of the model) for each
of the four effective dates of regulations listed in Table A-ll and for the
mean effective date of variances granted of December 1, 1973, using both the
additive and multiplicative exponential smoothing models. The null hypothesis
assumed in the Sign test was that the residuals had a probability destribution
with mean zero. The null hypothesis was rejected when the numbers of positive
and negative signs differed significantly from equality. The Runs test, in turn,
was used to test whether the positive and negative signs of the residuals were
distributed randomly in time or whether they were grouped in runs. A five per-
cent level of significance was used in both tests. The results are shown in
Table A-12 for the Sign test and in Table A-13 for the Runs test. Table A-12
shows the percentage of negative signs in each series of residuals and the cor-
responding percentage ranges indicative of significant changes in the trend of
measured air quality data. In Table A-13, a + sign indicates a significant up-
ward change, and a - sign indicates a significant downward change in the trend
of the measured air quality data coincident with the effective dates of fuel use
regulations or variances.
From Tables A-12 and A-13, it can be seen that the results
of the Sign and Runs tests are comparable. In general, the direction of trend
changes for each regulation or variance date is mixed; i.e., there are indica-
tions of both upward and downward changes depending on the data series analyzed.
The only definitive results are that a significant decrease in TSP levels is
noted coincident with the October 1, 1971, regulations (see Table A-ll). A down-
ward trend change for TSP and S02 levels on July 1, 1970, is noted, but unfortu-
nately, these results are based on data from only one monitoring station - Kenmore
A-64
Ulalden
-------
TABLE A-12
SIGN TEST: PERCENTAGE OF NEGATIVE SIGNS IN RESIDUALS
BETWEEN MEASURED AND PREDICTED AIR QUALITY LEVELS
FOR THE TIME PERIOD FOLLOWING THE ONSET OF
FUEL USE REGULATIONS AND VARIANCES
Regulations*
Jul. Oct. Oct.
Data Series 1970 1970 1971
Multiplicative Model
S02B 80 100
S02J No Data
S02L No Data
TSP1 100 0
TSP5 No Data
TSP7 No Data
Average 90 50
Additive Model
S02B 20 0
S02J No Data
S02L No Data
TSP1 40 0
TSP5 No Data
TSP7 No Data
Average 30 0
Trend Changes***
Downward if >_ 90 90
Upward if <_ 10 10
0
90
80
100
20
100
65
0
90
0
100
30
90
52
90
10
Jun.
1972
90
0
80
100
0
0
45
90
0
100
100
0
10
50
90
10
Variance**
Dec .
1973
100
0
83
100
33
83
67
67
0
100
100
33
67
61
100
0
*
Based on a 10-month time period.
**
Based on only a 6-month time period due to unavailability of data.
***
Percentage range indicative of significant changes in the trend of
measured air quality data at the 5% significance level.
fl-65 UhkJeni
-------
TABLE A-13
RUNS TEST: SIGNIFICANT CHANGES IN THE TREND OF MEASURED
AIR QUALITY DATA COINCIDENT WITH THE ONSET OF
FUEL USE REGULATIONS AND VARIANCES*
Regulations
Variance
Jul. Oct. Oct. Jun. Dec.
Data Series 1970 1970 1971 1972 1973
Multiplicative Model
S02B +
S02J ' No Data + +
S02L No Data
TSP1 +
TSP5 No Data +
TSP7 No Data - +
Additive Model
S02B
S02J
S02L
TSP1
TSP5
TSP7
+ +
No Data - + +
No Data +
+
No Data +
No Data - +
At 5% significance level, + denotes upward change in trend and
- denotes downward change in trend, while a blank denotes no
change in trend.
A-66
UJaldff
-------
Square, Boston. Occasional differences between the additive and multiplicative
results are due to the varying abilities of these two models in simulating time
series behavior.
Similarities in the patterns of the signs of the residuals
at different monitoring sites, produced by the above procedure, were compared in
order to judge if the measured series from several site locations qualitatively
exhibited similar behavior coincident with the onset of regulations or variances.
The appropriate statistical test of significance employed was the two-way analy-
sis of variance [18], with a 5% significance level. The results of the Sign test
analysis, shown in Table A-12, were used in these comparisons.
For each regulation or variance date, the column of per-
centages from Table A-12 was arranged in a three by four matrix representing the
three .representative monitoring sites (see Table A-2) and the four seasonally ad-
justed series analyzed from each site (viz., a multiplicative and additive model
of series components for both SC^ and TSP). The two-way analysis of variance
technique was then applied to these data to obtain estimates of the variance in
the data due to different monitoring sites and the residual variance, a measure
of the variance independent of that due to different sites or different data
series. The null hypothesis used was that the percentage of negative signs in
the data did not differ significantly from site to site, i.e., similarity. The
null hypothesis was rejected when an F-ratio of the variance due to different
sites and the residual variance indicated significant differences. This analysis
could not be applied to the July and October 1970 regulation dates due to the
lack of data from more than one monitoring site.
The null hypothesis was rejected (i.e., lack of similarity)
for every regulation and variance date except October 1971. These results in-
dicate that data from the three different monitoring sites exhibit qualitatively
similar behavior coincident with the onset of the October 1, 1971, fuel use regu-
lations.
b. Multiple Linear Regression Analysis
(1) Methodology
The utility of stepwise multiple regression was that it
permitted the identification of which of a large number of independent vari-
ables were most significant in explaining the variations in measured air,
A-67
Ulalden
-------
quality levels. The chosen variables are presented in the order of their
significance as predictors of the dependent variable and the multiple cor-
relation coefficient provides an indication of the overall fit of the mo-
del. The regression also provided an empirical functional relationship
between the dependent variable (ambient air quality levels) and the various
independent variables for use in future analyses.
The limitations that may be encountered in the appli-
cation of this technique include the definition of variables, i.e, func-
tional relations, and the testing of the statistical significance of the
derived regression equation. A standard assumption used in testing the
fit of a regression model to a given set of data is that the independent
variables be chosen at random from the set defined for the model. The
stepwise multiple regression procedure did not select predictors at ran-
dom, but instead chose them in the order of their significance in explain-
ing the variability in the data. For this reason, the standard F-test
for significance had to be modified prior to its application to the step-
wise regression results.
In order to illustrate how multiple regression analysis was
applied to the current problem, let the dependent variable, Y, represent
ambient S02 or TSP levels, and the independent variables X-|, X2, ••• XK
represent various periodic, meteorological, and regulatory predictors.
Assume that there are n sets of random observations of the Y and Xi vari-
ables. Multiple regression analysis attempted to fit this data to an
equation of the form:
Y = aO + alXl +a2X2+ •" +aKXK
where ag, a-|, ... aK are constant coefficients to be determined by the
regression analysis so as to provide a best fit to the data in the least
squares sense. The common measure of the fit of the model to the data was
provided by the multiple correlation coefficient, R. R2 is the amount of
variance of the dependent variable, Y, explained by the assumed regression
model.
/UUhj
-------
In the present study, a stepwise procedure was fol-
lowed in order to select the most significant variables from the total
set of those defined. The procedure consisted of adding one variable at
each step and computing the multiple correlation coefficient and the
residual sum of squares. At each step of the procedure, the variable
selected from the list of unused variables was the one that, when added
to the equation of the previous step, provided the greatest reduction in
the residual sum of squares. That is, the variable that provided the most
additional information about the dependent variable Y. Thus, at some
step in the procedure, a variable that was highly correlated with Y may
not have been chosen in favor of one that had a lower correlation with
Y. This occurred if the highly correlated variable simply duplicated in-
formation provided by a previously chosen variable. This masking of sig-
nificant variables (discussed in Section VI.A.2.B.(3)) occurred between
time and regulatory variables, and between seasonal and meteorological
variables in the current analysis.
In order to test that the variance in the data ex-
plained by the regression model was not merely a random occurrence, it
was necessary to determine the statistical significance of the fit pro-
vided by the model. For this purpose, an F-ratio test was used at each
step of the multiple regression procedure to determine whether the reduc-
tion in the residual sum of squares due to the added variable was statis-
tically significant [19]. The critical F value used was 3.0, i.e., an
independent variable was added into the regression if its F-ratio equalled
or exceeded 3.0. This critical value corresponds approximately to a ten
percent level of significance, i.e., there is at most a ten percent chance of
accepting a variable as significant in the regression when it actually
is not.
An important underlying assumption in application of
the F-test is that the added variable whose significance is being tested
has been randomly chosen for the total group defined. In the case of the
stepwise regression analysis, however, this choice was not random. In
fact, at each step the most significant of the remaining variables was
A-69
Maiden
-------
selected, and therefore, the F-test was likely to overestimate the sig-
nificance of the added variable. The use of a critical F value of 3.0
may correspond to an actual level of significance higher than ten percent
in any one given multiple regression. However, the regression analysis
was performed on not one, but twenty data series representing thirteen
different monitoring sites in Metropolitan Boston (see Table A-l), and
the obtained results from all of these regressions were combined in order
to make general statements regarding the effects of fuel use regulations
and variances on ambient air quality levels. The combined or overall
level of significance was, therefore, smaller than any one individual
level of significance [20]. For example, if one has K independent tests,
each significant at level Pj, i = 1 K, then the overall level of
significance is found by use of the chi square distribution where -2.E £nP.j
has a chi square distribution with 2K degrees of freedom. Using this
expression, it can be shown that if we have two tests, each significant at
the ten percent level, the overall significance is smaller at approxi-
mately five percent. Thus, when combining the results of many tests, each
using a critical value of 3.0, it can be said with certainty that the
overall level of significance is equal to or lower than five percent.
The independent variables selected for the regres-
sion analysis represented meteorological conditions, seasonal variations,
regulatory impositions and time. Four dummy variables [19] quantifying
the effects of the four significant regulatory dates listed in Table A-ll
were chosen along with a dummy variable for the mean effective date of
special variances granted of December 1, 1973. These variables were de-
fined as 0 before the associated effective date and 1 otherwise. Twelve
dummy variables quantifying the effect of seasonal variations were used
in the regression analysis. Each monthly variable was defined as 1 on its
associated month and 0 otherwise. Additional independent variables were
monthly heating degree-days and the reciprocal of monthly average wind
speed, both measured at Boston's Logan International Airport, and time
variables, t, t2, and t3, associated with the number of months (t) since
the beginning of the measured S02 or TSP data time series.
A-70
Walden
-------
The dependent variable in the regression analysis
consisted of either the measured ambient S02 and TSP levels or the nat-
ural logarithms of these quantities. The results obtained by analyzing
the data and the logarithms of the data were so similar that only those
associated with analysis on the original data are presented in this re-
port. The regression analysis was applied separately to each of the
twenty available time series shown in Table A-l.
(2) Zero Order Correlations
The zero order (i.e., simple) correlations between
the dependent variables and various independent variables are shown in
Tables A-14 and A-l5, respectively, for the S02 and TSP data. The sign
of the correlation coefficients indicates whether these variables are
positively or negatively correlated. An absolute value of 0.31 or larger
for these simple correlation coefficients corresponds to the five percent
significance level, i.e., there is at least a 95 percent chance that a
correlation is significant. The cut-off value of 0.31 is conservative
in that it is based on the smallest sample size of all the time series
analyzed.
From Tables A-14 and A-l5, it can be seen that the
regulation and variance variables are consistently negatively correlated
with ambient S02 and TSP levels. Statistical significance is attained
most often for the regulations effective in July and October 1970 and
October 1971. The correct interpretation of these negative correlations
is that the average S02 and TSP levels after an effective date of a regu-
lation or variance are statistically significantly lower than the average
levels before this date. The large quantity of significant negative cor-
relations indicates we are justified in further analysis of the relation-
ship between changes in the S02 and TSP levels and the specific regula-
tions and variances.
A-71
lUUen.
-------
TABLE A-14
**
CORRELATION COEFFICIENTS BETWEEN AMBIENT S02 LEVELS AND VARIOUS OTHER VARIABLES
AT SEVERAL MONITORING SITES IN METROPOLITAN BOSTON
Variable SO
Reciprocal of
Mean Wind Speed
Degree Days
Regulations:
July 1, 1970
Oct. 1, 1970
Oct. 1, 1971
June 1, 1972
Variance:
Dec. 1, 1973
Time (t)
* No data available
** An absolute value
significance level
2A
406
485
538
485
508
412
179
566
of 0.
S02B
-.430
.535
-.637
-.561
-.601
-.589
-.286
-.748
31 or
so2c
-.379
.424
*
*
-.509
-.443
-.261
-.607
larger for
S02D
-.344
.360
-.492
-.438
-.557
-.590
-.284
-.711
these
S02E
-.399
.461
*
*
-.390
-.342
-.135
-.423
Data
S02F
-.263
.163
*
*
-.268
-.358
.023
-.338
simple correlation
S e
S02G
-.276
.248
*
*
-.433
-.185
-.183
-.373
r i e s
S02H
-.295
.347
*
*
-.358
-.426
-.063
-.474
coefficients
so2i so2J
-.402 -
.469
*
*
-.232 -
-.260 -
.009
-.306 -
corresponds
.559
.679
*
*
.291
.200
.047
.299
to the
S02K
-.448
.607
*
it
-.053
-.065
-.148
-.206
five
S02L
-.299
.455
*
*
-.220
-.325
.051
-.220
percent
S02M
-.358
.341
*
*
-.278
-.175
.103
-.196
-------
--J
CO
TABLE "A-15
**
CORRELATION COEFFICIENTS BETWEEN AMBIENT TSP LEVELS AND VARIOUS OTHER VARIABLES
AT SEVERAL MONITORING SITES IN METROPOLITAN BOSTON
Variable
Reciprocal of
Mean Wind Speed
Degree Days
Regulations:
July 1, 1970
Oct. 1, 1970
Oct. 1, 1971
June 1, 1972
Variance:
Dec. 1, 1973
Time (t)
TSP 1
-.291
.500
-.716
-.550
-.403
-.497
-.262
-.593
TSP 2
- .072
-.069
*
*
-.448
-.298
-.142
-.355
Data
TSP 3
.293
-.190
*
*
-.301
-.044
-.258
-.166
S e r
TSP 4
.483
-.359
*
*
-.587
-.484
-.431
-.550
i e s
TSP 5
-.056
.154
*
*
-.044
.074
-.134
-.052
TSP 6
-.080
.226
*
*
-.572
-.445
-.224
-.567
TSP 7
.656
-.555
*
*
-.423
-.079
-.399
-.272
* No data available
** An absolute value of 0.31 or larger for these simple correlation coeffi-
cients corresponds to the five percent significance level.
-------
(3) Partial Correlation Structure
The results in Table A-14 show that degree-days are signifi-
cantly positively correlated with S02 levels at most monitoring stations. That
is, both quantities vary seasonally, with winter maxima and summer minima. From
the meteorological observations, a strong correlation (-0.743) was found between
heating degree-days and the reciprocal of mean wind speed for Boston, reflecting
the climatological relationship between these quantities. The implication of this
interdependence is that partial correlations of S02 levels with the reciprocal of
wind speed, removing the linear relationship with degree-days, are all close to
zero. In other words, the previously discussed masking situation is occurring
where, if degre.e-days are already present in the regression, they dominate the
correlations, and the addition of wind speed as a variable adds little or no
information about the S02 levels. Thus, the apparent negative correlation be-
tween SOo levels and the reciprocal of wind speed in Table A-14 is due princi-
pally to the strong correlation climatologically between wind speed and degree-days.
Table A-15 shows that the same situation is observed for the
TSP1 (Kenmore Square, Boston) series. Of the remaining TSP series, only TSP4
(Dedham Avenue, Needham) and TSP7 (Beaver Street, Waltham) have significant cor-
relations with both wind speed and degree-days. At these sites, where wind speed
is the more important of the two variables, the correlation of the reciprocal
with TSP levels is positive, i.e., both quantities vary seasonally with maxima
in the summer and minima in the winter months. The partial correlations of TSP
levels with degree-days at these two sites, removing the linear relationship with
wind speed, are all close to zero, emphasizing the dominant effect of wind speed
on TSP levels.
In order to investigate the possibility that statis-
tically significant downward changes in S02 and TSP levels coincident with
the effective dates of fuel use regulations and variances are due princi-
pally to changes in meteorological conditions, the partial correlations of
S02 or TSP levels with the regulatory and variance variables were computed,
removing the linear relationship with either degree-days or the reciprocal
of wind speed. The analysis was conducted for the six representative time
series (see Table A-2), and the results are shown in Table A-16. It can
be seen that all the correlations are still negative, and the majority are
still statistically significant. Note that some correlations which were
A-74
llUakkni
-------
I
•vj
Ul
TABLE A-16
PARTIAL CORRELATION COEFFICIENTS** OF AMBIENT S02 OR TSP LEVELS
WITH REGULATORY OR VARIANCE VARIABLES
AT THREE REPRESENTATIVE MONITORING SITES
(Removing the Linear Relationship with either Degree-Days or Wind Speed)
Reciprocal of Wind Speed Removed
Variable S02B Sty S02L TSP 1 TSP 5 TSP 7
Degree- Days Removed
S02B S02J S02L TSP 1 TSP 5 TSP 7
Regulations:
July
Oct.
Oct.
June
1, 1970 -.70 * * -.74 * *
1, 1970 -.62 * * -.57 * *
1, 1971 -.61 -.29 -.24 -.39 -.04 -.55
1, 1972 -.59 -.16 -.33 -.48 -.08 -.15
-.69 * * -.76 * *
-.72 * * -.68 * *
-.74 -.44 -.38 -.48 -.05 -.36
-.64 -.14 -.34 -.51 -.10 -.13
Variance:
Dec. 1, 1973 -.46 -.15 -.06 -.37 -.16 -.25
-.48 -.17 -.10 -.43 -.18 -.29
* No data available
** An absolute value of 0.32 larger for these correlation coefficients corresponds to the five percent
significance level. The value increases because of a loss of one degree of freedom in the partial
correlation coefficients.
-------
significant are no longer (e.g., TSP7 and the variance), while others
which were not significant before now are (e.g., S02B and the variance).
In general, the pattern of correlations in Table A-16 is the same as that
shown in Tables A-14 and A-15. Thus, it appears that statistically sig-
nificant decreases in average S02 and TSP levels are associated with the
effective dates of regulations and variances even after meteorological
conditions have been taken into account.
(4) Stepwise Multiple Linear Regression
Stepwise multiple linear regression analysis was ap-
plied to the 13 S02 and seven TSP time series (see Table A-l). All pre-
viously listed independent variables were used except those involving
time which would mask the effects of other independent variables. Tables
A-l7 and A-18 show, respectively, those variables which, when added to
the regression in Stepwise fashion, were significant in explaining the
variance of the measured S02 and TSP time series at the 10 percent sig-
nificance level. To aid in ranking the importance of individual indepen-
dent variables in the final regression, the beta weights and t-statistics
of the final regression are included in Tables A-17 and A-18, along with
the multiple correlation coefficients, R. The final regression is the
regression that includes all the variables that the Stepwise procedure
accepted as significant for inclusion. In general, the larger the beta
weight (in absolute value), the greater the total contribution the inde-
pendent variable makes to the final regression. The sign of the beta
weight indicates whether the associated independent variable is negatively
or positively correlated with the dependent variable, given that linear
relationships with all other independent variables of the final regres-
sion have been removed. The t-statistic tests the hypothesis that the
corresponding independent variable makes a significant added contribution
to the final regression given that the contributions of all other indepen-
dent variables of the final regression have been taken into account. A
value of the t-statistic greater than or equal to 2 in absolute value in-
dicates significance at the five percent level. The results of the beta
weight and t-statistics would be the same if the independent variables in
A-76
-------
TABLE A-17
STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF MEASURED S02 LEVELS
Step
Number
Added Variable
Final
R Beta ^Weight
Final
t Statistics
Government Center, Boston (S02A)
1
2
3
4
5
6 v
Regulation:
Degree Days
Regulation:
Seasonal :
Variance:
Seasonal :
July 1970
October 1971
February
December 1973
January
.538
.718
.804
.820
.835
.847
.353
.376
.355
.235
.169
.172
-4.059
3.653
-3.981
2.562
-2.016
1.848
Kenmore Square, Boston (S02B)
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
Regulation:
Degree Days
Regulation:
Variance:
Regulation:
Regulation:
Degree Days
Seasonal :
Variance:
Seasonal :
Central
Regulation:
Regulation:
Degree Days
•Regulation:
Variance:
July 1970
October 1971
December 1973
October 1970
South Bay,
October 1971
'
October
December 1973*
March
Square, East
June 1972
July 1970
October 1971
December 1973
.637
• .788
.888
.916
.922
Boston (S02C)
.509
.699
.768
.799
.822
Boston (S02D)
.590
.676
,751
.769
.787
.236
.589
.350
.245
.185
.519
.622
.290
.226
.199
.200
.289
.415
.280
.187
.-2.448
9.717
-4.984
-4.058
-1.789
-5.694
6.634
3.241
-2.403
-2.180
-1.385
-2.892
4.268
-1.948
-1.819
A-77
Ulalden
-------
TABLE A-il7i (continued)
STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF MEASURED S02 LEVELS
Step
Number Added Variable
Greenough Street
1 Degree Days
Final
R Beta Weight
, Brookline (S02E)
.461 .482
2 Regulation: October 1971 .619 -.414
Village Street,
Marblehead (S02F)
1 Regulation: October 1971 .434 -.374
2 Regulation: June 1972
U.S. Army Site
1 Regulation: October 1
2 Degree Days
Main Street,
1 Regulation: June 1972
2 Degree Days
Dedham Avenue
1 Degree Days
2 Regulation: October 1
3 Seasonal : February
Nahatan Street
1 Degree Days
2 Regulation: October 1
.511 -.277
, Maynard (S02G)
971 .433 -.523
.565 .375
Medford (S02H)
.426 -.383
.513 .290
, Needham (S02D
.469 .421
971 .588 -.378
.665 .340
, Norwood (S02J)
.679 .695
971 .753 -.326
Final
t Statistics
3.830
-3.291
-2.545
-1.889
-3.744
2.679
-2.756
2.092
3.018
-2.945
2.488
6.589
-3.089
Hancock Street, Quincy (S02K)
1 Degree Days
2 Variance: December
3 Seasonal : November
.607 .655
1973 .679 -.282
.711 .213
5.449
-2.345
1.813
A-78
-------
TABLE A-17 (continued)
STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF MEASURED S02 LEVELS
Step
Number
Added Variable
R
Final
Beta Weight
Final
t Statistics
Beaver Street, Wai than (S02L)
1 Degree Days .455 .540 3.868
2 Regulation: October 1971 .568 -.360 -2.513
Montvale Avenue, Woburn (SOJ1)
1 Seasonal: March .375 .312 2.220
2 Reciprocal of Wind Speed .472 -.270 -1.914
3 Regulation: October 1971 .530 -.242 -1.753
A'79 HI U
Ulaldeni
-------
TABLE A-18
STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF MEASURED TSP LEVELS
Step
Number Added Variable
Kenmore Square, Boston
1 Regulation:
2 Degree Days
3 Variance:
4 Seasonal :
5 Seasonal :
6 Regulation:
7 Seasonal:
July 1970
.
December 1973
November
December
June 1972
April
Greenough Street, Brookl
1 Regulation:
2 Seasonal :
U.S
1 Regulation:
2 Seasonal :
3 Reciprocal
4 Regulation:
5 Seasonal :
6 Variance:
October 1971
May
Final
R Beta Weight
(TSP 1 )
716
828
873
901
916
926
932
ine (TSP
448
599
-.517
.548
-.267
-.233
-.163
-.153
.108
2)
-.422
.398
Final
t Statistic
-8.920
9.125
-4.357
-4.218
-2.830
-2.480
1.984
-3.200
3.012
. Army Site, Maynard (TSP 3)
October 1971
February
of Wind Speed
June 1972
September
December 1973
301
431
586
648
686
729
-.617
.446
.318
.484
-.258
-.285
-4.069
3.547
2.390
3.058
-2.113
-2.083
Dedham Avenue, Needham (TSP 4)
1 Regulation:
2 Reciprocal
3 Regulation:
4 Seasonal :
5 Seasonal :
6 Seasonal :
7 Seasonal :
8 Seasonal :
October 1971
of Wind Speed
July 1973*
July
October
November
February
March
587
754
793
822
848
873
886
898
-.505
.680
-.239
-.286
-.196
-.164
.192
.155
-5.826
6.848
-2.869
-3.129
-2.401
-1.967
2.211
1.838
A-80
-------
TABLE A-18 (continued)
STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF MEASURED TSP LEVELS
Step Final Final
Number Added Variable R Beta Weight t Statistic
Nahatan Street, Norwood (TSP 5)
1 Seasonal: February .507 .506 3.715
Hancock Street, Quincy (TSP
1
2
3
1
2
3
4
5
6
7
8
Regulation: October 1971
Seasonal : January
Seasonal : June
Beaver Street, Wai
Reciprocal of Wind Speed
Regulation: October 1971
Seasonal : July
Regulation: June 1972
Regulation: July 1973 *
Seasonal: April
Degree Days
Seasonal : June
.572
.711
.762
tham (TSP
.656
.775
.837
.860
.894
.913
.923
.933
6)
-.540
.448
.274
7)
.974
-.665
-.336
.435
-.268
.218
.272
.156
-5.055
4.186
2.554
8.800
-7.266
-4.304
4.612
-3.565
31281
2.500
2.132
* A regulatory variable for July 1, 1973 was included as an independent var-
iable in the original regression analysis before it was learned the regula-
tions were in fact never implemented. As a point of interest, this variable
was never chosen once in the stepwise regression analysis of S02 levels and
only twice in the analysis of TSP levels. No statistically significant and
consistent changes in TSP levels are associated with this date.
A-81
UJalden
-------
the regression were statistically independent, i.e., if zero-order cor-
relation coefficients between them were nearly zero. This was not, how-
ever, the case as significant interdependence existed between seasonal
and meteorological variables, and between time and regulatory variables.
To aid in extracting information from Tables A-17
and A-18, the regulations and variances which proved to be significant
for each of the analyzed S02 and TSP time series (i.e., those chosen by
the stepwise procedure), are indicated in Tables A-19 and A-20, respec-
tively. Here, a + sign indicates significant increase and a - sign in-
dicates a significant decrease in ambient S02 and TSP levels after the
associated regulation or variance date.
Conclusions based on the stepwise regression analysis
indicate that:
(a) The October 1971 Regulation (see Table A-ll)
was associated with a significant decrease
in S02 and TSP levels in 80% of the data
series.
(b) The July 1970 Regulation is also associated
with a significant decrease in S02 and TSP
levels, this time in 100% of the data series.
However, this conclusion is based on data
from only three monitoring stations for the
S02 and one monitoring station for the TSP
levels.
(c) No statistically significant rise in S02 and
TSP concentrations occurred at any of the
sites coincident with variances granted, be-
cause of the energy shortage, during the winter
of 1973-1974. In fact, a statistically sig-
nificant decrease in levels of these pollu-
tants during this winter occurred in 35% of
the data series, commencing in December 1973.
(5) Complete Multiple Linear Regression
In addition to the stepwise procedure, a complete
multiple linear regression analysis was performed on all twenty data
series. Here, all independent variables except time were included in the
A-82
UJaldff
-------
TABLE A-19
STATISTICALLY SIGNIFICANT REGULATIONS AND VARIANCE
IN THE STEPWISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF AMBIENT S02 LEVELS*
I
00
co
October 1971
June 1972
Variance
December 1973
S02A S02B S02C S02D
Data Series Analyzed
Regulations
July 1970
October 1970
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND denotes no data
* At 10% significance level,
+ denotes Increase .in measured levels
- denotes decrease in measured levels
a blank denotes no change in measured levels
-------
TABLE A-20
STATISTICALLY SIGNIFICANT REGULATIONS AND VARIANCE
IN THE STEPNISE MULTIPLE LINEAR REGRESSION ANALYSIS
OF AMBIENT TSP LEVELS*
:&
oo
TSP 1
Data Series Analyzed
TSP 2 TSP 3 TSP 4 TSP 5 TSP 6 TSP 7
Regulations
July 1970
October 1970
October 1971
June 1972
ND
ND
-
_
ND
ND
-
+
ND
ND
-
ND ND
ND ND
-
ND
ND
-
+
Variance
December 1973
ND denotes no data
* At 10% significance level,
+ denotes increase in measured levels
- denotes decrease in measured levels
a blank denotes no change in measured levels
-------
regression rather than just those with an F ratio above the critical value
of 3.0. Tables A-21 and A-22 give the results of these regressions in the
format used in Tables A-19 and A-20. Here, a regulation was noted as sig-
nificant only if the corresponding t-statistic in the complete regression
was significant. That is, if the added contribution of the regulatory
variable, after all other independent variables have been taken into ac-
count, was significant at the five percent level. This was clearly a
very stringent test. It can be seen from Tables A-21 and A-22 that only
the October 1971 regulation was still strongly significant in both the
502 and TSP analyses after application of this new test. These analyses
seem to be clearly demonstrating the effectiveness of that regulation.
Note that the variance variable was not significant in the analysis of
any of these series using this stringent test and so was not included in
Tables A-21 and A-22.
Another interesting analysis was performed on the
results of the complete regressions. The signs of the regression coef-
ficients for the regulatory and variance variables (regardless of signi-
ficance) were summarized from the complete regressions in Table A-23. The
sign of a regression coefficient indicates the direction of change in the
pollutant time series due to the independent variable involved. A nega-
tive sign indicates a drop in the series after the associated regulatory
variance data and a positive sign represents a rise. A Sign test [25]
(at the 5% level of significance) applied to the data in Table A-23, in-
dicates statistically significant decreases in S02 and TSP levels occurred
regionally after the October 1971 and December 1973 dates. This analysis
takes the effects of all other variables into account before assessing
the relationship with the regulatory or variance variable. These results
are important in that they represent a combined analysis for all sites
and the validity of the analysis does not depend on the usual regression
analysis assumptions, e.g., no serial correlation (autocorrelation) in
the time series data.
Utiden,
-------
Ja
CO
TABLE A-21
STATISTICALLY SIGNIFICANT REGULATIONS IN THE COMPLETE
MULTIPLE LINEAR REGRESSION ANALYSIS
OF AMBIENT S02 LEVELS*
Regulations
July 1970
October 1970
October 1971
June 1972
Data Series
S02A S02B S02C S02D S02E S02F
ND ND ND
ND ND ND
Analyzed
S02G S02H S02I S02J S02K S02L S02M
ND ND ND ND ND ND ND
ND ND ND ND ND ND ND
+ +
At 5% significance levels (t statistic for complete regression)
ND Denotes no data
+ Denotes increase in measured levels
- Denotes decrease in measured levels
a blank denotes no change in measured levels
-------
TABLE A-22
STATISTICALLY SIGNIFICANT REGULATIONS IN THE COMPLETE
MULTIPLE LINEAR REGRESSION ANALYSIS
OF AMBIENT TSP LEVELS*
Data Series Analyzed
Regulations TSP1
July 1970
October 1970
TSP2
ND
ND
TSP3
ND
ND
TSP4
ND
ND
TSP5
ND
ND
TSP6
ND
ND
TSP7
ND
ND
October 1971
June 1972
At 5% significance level (t statistic for complete regression)
ND Denotes no data
+ Denotes increase in measured levels
- Denotes decrease in measured levels
a blank denotes no change in measured levels
A-87
Maiden
-------
TABLE A-23
SIGNS OF REGRESSION COEFFICIENT IN COMPLETE REGRESSION
Series
Name
S02A
B
C
D
E
F
G
H
I
J
K
L
M
TSP 1
2
3
4
5
6
7
Regulatory and Variance Variables
July Oct. Oct. June Dec.
1970 1970 1971 1972 1973
_
-
ND
-
ND
ND
ND
ND
ND
ND
ND
ND
ND
-
ND
ND
ND
ND
ND
ND
_ —
-
ND
-
ND
ND
ND
ND
ND
ND
ND
ND
ND
+
ND
ND
ND
ND
ND
ND
+
-
+
+
+
+
+
+
+
+
+
-
+
-
+
+
-
+
-
+
ND denotes data not available
A-88 U/aldenl
-------
(6) The Independent Variable Time
One possible danger in the proceeding analysis was
that the downward trends in ambient S02 and TSP levels that were signi-
ficantly correlated with regulatory and variance information may have
been already occurring, due to other causes, prior to the beginning of
available data in 1970. This concern was addressed by redoing the step-
wise multiple regression analysis with the time variables, t, t2 and t3,
added into the pool of independent variables. This analysis was undertaken,
and of the twenty regressions shown in Tables A-17 and A-18, only six
changed. Tables A-24 and A-25 present the results of the new computa-
tions, presented in tie same format used in Tables A-19 and A-20.
The important changes occurred in series S02A, S02B,
S02C and S02D. The reasons these series showed significant correlation
with the time variables was probably due to the fact that these series had
the highest S02 levels prior to the first fuel use regulation in July
1970 and that each subsequent regulation reduced the series somewhat. In
a deseasonalized (i.e., seasonal component removed) and error-free series,
the combined effects of several succeeding regulations would produce a
consistently decreasing trend that could appear to be due to an inverse
relationship with increasing time. Supporting this hypothesis was the
fact that the October 1971 regulation was no longer associated with a
significant decrease in measured S02 levels in series S02A, S02B, S02C,
and S02D in Table A-20 as it was in Table A-19 and A-21. That is, once
time entered the stepwise regression as an in independent variable, it
masked any correlations between regulatory or variance variables and mea-
sured air quality levels. Based on available data, it was therefore con-
cluded that decreases in measured S02 and TSP levels concurrent with the
effective dates of fuel use regulations were not due to a pre-existing
downward trend in the data.
A-89 lUaUen,
-------
>
o
July 1970
October 1970
October 1971
June 1972
Variance
December 1973
TABLE A-24
STATISTICALLY SIGNIFICANT REGULATIONS AND VARIANCE
IN A STEPWISE MULTIPLE LINEAR REGRESSION
ANALYSIS OF AMBIENT S02 LEVELS*
(TIME VARIABLES INCLUDED)
Time Series Analyzed
Regulations S02A S02B S02C S02D S02E S02F S02G S02H S02I S02J S02K S02L
Time
Variables
ND
ND
+
+
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
ND
At 5% significance level
ND Denotes no data
+ Denotes increase in measured levels
- Denotes decrease in measured levels
a blank denotes no change in measured levels
-------
TABLE A-25
STATISTICALLY SIGNIFICANT REGULATIONS AND VARIANCE
IN A STEPNISE MULTIPLE LINEAR REGRESSION
ANALYSIS OF AMBIENT TSP LEVELS*
(TIME VARIABLES INCLUDED)
Time Series Analyzed
Regulations
July 1970
October 1970
TSP1 TSP2
ND
ND
TSP3
ND
ND
TSP4
ND
ND
TSP5
ND
ND
TSP6
ND
ND
TSP7
ND
ND
October 1971
June 1972
Variance
December 1973
Time
Variables
t
t2
t3
At 5% significance level
ND Denotes no data
+ Denotes increase in measured levels
- Denotes decrease in measured levels
a blank denotes no change in measured levels
A-91
Walden
-------
(7) Lagged Effects of Regulations and Variances
The regulatory and variance variables used in the
preceeding regression analyses assumed that the implementation and sub-
sequent effects of the corresponding regulation or variance took place
simultaneously with its imposition. It may have been the case, however,
that there was a lagged effect due to an implementation delay on the
part of some sources. For example, a fuel use regulation limiting the
sulfur content of certain fuels burned in the area that became law in
October may not have shown itself in S02 levels until November (one-month
lag) or even until December (two-month lag). It is further possible
that there was a partial effect the first month, a further effect the
second month and finally the full effect by three months. Regression
analyses for these three situations were performed on the six represen-
tative series (see Table A-2). In no case was there a statistically sig-
nificant gain over the methods originally employed for representing the
regulatory and variance variables. Note however, it is entirely possible
that the monthly data used in the statistical data analyses were too crude
to detect any real lag effect.
(8) Test for Autocorrelation
An assumption made in applying statistical tests to
the results of the stepwise regression analysis was that the residuals of
the regression did not possess any autocorrelation, i.e., there was no un-
explained serial correlation in the data. The validity of the statistical
tests depended on this assumption. To ensure validity, the Runs test [18]
was applied to the residuals from the stepwise regressions of the six
representative series. Of these six only the S02B series (Kenmore Square)
was found to possess significant autocorrelation, at the five percent
significance level. This can be considered just a chance occurrence,
i.e., in six series there is a good chance one will produce significance
at the five percent level even if all six are uncorrelated. Another reason for
finding significant autocorrelation in the Kenmore Square series could be that
this series has a larger number of daily S02 measurements per month than any of
the other series. One would expect values taken serially to exhibit serial cor-
relation. Thus, the application of statistical tests to the results of the step-
wise regressions was valid. Any further analysis of the data should entail an
A-92 uum
-------
examination of the autocorrelation functions of the time series, possibly
using the techniques of Box and Jenkins [21]. However, the time series
may not be long enough for these types of statistical tests.
(9) Results
The results of the statistical regression analyses
indicate that major SIP fuel use regulations effective July 1, 1970 and
October 1, 1971 (see Table A-ll) are associated with statistically sig-
nificant decreases in regional S02 and TSP levels. This is explained by
the fact that both of these regulations were responsible for the conver-
sion of many large fuel utilization facilities to fuels with lower sulfur
and ash contents. The first set of regulations, limiting the ash content
of fossil fuels burned in the AQCR, caused conversions principally from
coal to oil and gas. Note that although this set of regulations did not
limit the sulfur content of fuels, the net result of these fuel conver-
sions was usage of fuels with a lower ash and sulfur content. The October
1, 1971 regulations, severely restricting the sulfur content of residual
oil burned in the AQCR, caused suppliers of residual oil to blend cleaner
distillate oil into non-complying residual in order to produce fuel stocks
of acceptable quality.
Additional results show that no statistically sig-
nificant rise in regional S02 and TSP concentrations occurred coincident
with variances granted, because of the energy shortage, during the winter
of 1973-1974. In fact, the results of the complete regressions indicate
statistically significant decreases in these pollutant levels occurred
regionally commencing in December 1973. The implication of this finding
is that some other mechanism, most probably fuel conservation efforts by
consumers, overrode the effects of any increase in S02 and TSP emissions
due to the combustion of nonconforming fuel.
The following table indicates the variables and events
which, in the stepwise regression, made the greatest total contribution to-
ward explaining variations in regional S02 and TSP levels between 1970 and
1974. These quantities are listed in the order of their ability to explain
these variations.
A-« lUJalden
-------
S02 TSP
(Entire Region) (Urban Core) (Non-Urban)
Heating Degree-Days Heating Degree-Days (Mean Wind Speed)"
October 1, 1971, July 1, 1970, Regu- October 1, 1971,
Regulations 1 ations Regulations
In the cases where heating degree-days and the reciprocal
of mean wind speed were the most significant variables in the multiple linear
regression, they were found to be directly proportional to measured pollutant
concentrations. Degree-days are a measure of local fuel burning activities
and emissions; mean wind speed is a measure of the dilution capability of the
atmosphere.- These proportionality results are consistent with the relation-
ships normally assumed between pollutant concentrations, emissions, and wind
speed in most air quality models. Because of the strong negative correlation
climatologically between the reciprocal of wind speed and degree-days, in cases
where degree-days were found to be the most important variable (S02 and urban
core TSP measurements), the correlation between pollutant concentrations and
the reciprocal of wind speed was necessarily negative. Removing first the
relationship between pollutant concentrations and degree-days, essentially
zero correlation was found between the pollutant concentrations and the re-
ciprocal of wind speed. This illustrates the type of illusory results that
can be obtained when there is significant intercorrelation among variables.
The conclusions based on the above results are: (1) fuel
burning emissions dominate S02 concentrations throughout the Metropolitan Bos-
ton AQCR; (2) TSP concentrations in the urban core are also dominated by fuel
burning sources; but (3) TSP levels at non-urban sites are dominated by emission
sources other than fuel burning facilities, most probably local particulate
sources, road dust, and pollen. These results confirm the classical seasonal
patterns and their urban/non-urban split found in the measured S02 and TSP
data in the data analysis task.
c. Trends in Measured Air Quality Data during the Period of Variances
Results from the X-ll ratio to moving average data ana-
lysis were interpreted for trends during the period of variances. Two
indicators of the trend of the measured air quality data, the Trend-Cycle
(T) and the Months for Cyclical Dominance (MCD) components, are shown in
Tables A-26 and A-27 for all data series for the time period during which
variances were granted on fuel use regulations in the Metropolitan BostoE
UhUenl
-------
TABLE A-26
Monthly Average S02 Concentrations Measured During the Period December, 1973
to June, 1974 in the Metropolitan Boston Air Pollution Control District*
Data
S02A
S02B
S02C
S02D
S02E
S02F
S02G
S02H
S02I
S02J
S02K
S02L
S02M
*
Trend
Dec.
Series 1973
20.6
22.4
18.9
9.0
16.9
15.8
6.0
21.0
8.9
19.0
18.9
10.6
20.5
In micrograms per cubic
Cycle Series
May
1974
25.5
24.5
8.7
9.6
13.6
13.9
4.2
16.1
9.1
12.6
15.1
17.4
11.3
meter
June
1974
29.5
-
6.5
-
12.4
12.8
3.7
14.8
9.0
10.2
-
19.0
8.0
MCD** Series
Dec.
1973
23.3
23.5
18.0
10.0
17.7
17.4
6.3
28.5
9.2
16.5
18.4
10.4
19.2
March
1974
18.3
20.6
13.4
10.5
13.8
16.0
4.4
24.6
9.8
16.4
17.6
10.9
17.7
Apri 1
1974
22.5
-
13.4
-
12.4
13.8
4.6
23.9
10.1
17.4
-
14.6
16.6
** Months for cyclical dominance
Ulabkn
-------
TABLE A-27
Monthly Average TSP Concentrations Measured During the Period December, 1973
to June, 1974 in the Metropolitan Boston Air Pollution Control District*
Data Series
TSP1
TSP2
TSP3
TSP4
TSP5
TSP6
TSP7
Trend
Dec.
1973
87.6
41.0
24.4
32.4
37.7
48.0
44.0
Cycle
May
1974
80.9
36.8
19.0
25.3
41.4
44.1
27.7
Series
June
1974
84.7
38.1
18.4
24.2
41.9
25.0
Dec.
1973
89.4
38.6
24.9
32.5
39.7
51.2
44.7
MCD Series
March
1974
74.3
36.4
20.8
28.1
38.6
44.6
33.6
April
1974
78.4
36.6
20.1
28.7
40.2
26.0
* In micrograms per cubic meter
** Months for cyclical dominance
A-96
Maiden
-------
AQCR, i.e., December 1973 through June 1974. These component series were pro-
duced using a multiplicative model in the X-ll ratio to moving average analysis.
Note that due to the manner in which the MCD component series are computed, they
end two months before the trend-cycle series.
A comparison of the December 1973 values with later values
indicates quite clearly that the ambient levels of S02 and TSP did not
rise in general during the period December 1973 to June 1974. On the
contrary, significant decreases in ambient S02 and TSP levels were found
to occur during this period using the Sign test at the 5 percent signi-
ficance level applied to the MCD component data. As outlined in Section
VI.A.l.a.(l), the MCD series is the most sensitive indicator output by
the X-ll analysis for judging the behavior of short-term trends.
d. Trends in Measured Heating Degree-Day and Wind Speed
Data
Previous analysis employing exponential smoothing and
multiple linear regression techniques focused on the time periods asso-
ciated with the effective dates of fuel use regulations and variances in
order to observe any concurrent changes in the trend of measured air qual-
ity data. Additional time periods which warranted investigation were
those during which heating degree-days and mean wind speed deviated sig-
nificantly from climatological averages. Degree-days are a measure of
local fuel burning activities and emissions; mean wind speed is a mea-
sure of the dilution capability of the atmosphere.
Monthly heating degree-days and mean wind speed measured
at Boston's Logan International Airport are shown in Figures A-41. and A-42 ,
respectively, as a ratio to the 30-year measured climatological averages.
Gaps in the plot shown in Figure A-47 correspond to the summer season dur-
ing which the ratio becomes meaningless due to diminishingly small or
zero monthly degree-days totals. For this reason, extreme values which
occurred in the late spring and early fall of several years, proximate
to the summer season, had to be ignored. A significance level of 20% de-
viation -from climatology was arbitrarily chosen and marked off on the
graphs.
A-97 Iiu u
ilUalden
-------
I
oo
Figure A-40
Ratios of Monthly Heating Degree Days to 30-year Climatological Averages in Boston, Mass.
+ 20%
- 20%
Jan
1970
Jan
1971
Jan
1972
Jan
1973
Jan
1974
-------
Figure A-41
Ratios of Monthly Mean Wind Speed to 30-year Climatological Averages in Boston, Mass.
20%
- 20%
1974
-------
It can be seen from Figure A-41 that the significant
features are a marked increase in heating degree-day ratios from Septem-
ber 1971 to May 1972 and a decrease from October 1972 to March 1973. An
examination of the trend-cycle component graphs for the six representative
data series (Figures A-2 through A-7) showed that concurrent changes in
the trend of measured air quality data occurred in the S02B (Kenmore Square,
Boston) and TSP7 (Beaver Street, Waltham) series. During these two time
periods, the trend-cycle component of the S02B series is positively cor-
related with heating degree-day totals. The trend-cycle component of the
TSP7 series, however, is negatively correlated with heating degree-days
in the same time periods, moving in the opposite directions. These re-
sults are in agreement with corresponding statistics derived in the mul-
tiple regression analysis and shown in Tables A-14 and A-15.
The graph of mean wind speed ratios (Figure A-42) exhibits
only one significant feature, a marked decrease from April through October
1973 with a low point in July 1973. Concurrent rises in trend-cycle com-
ponents during this period with peaks near July 1973 occurred in the S02B,
TSP1 (Kenmore Square, Boston), and TSP7 series. These results compare
favorably with the correlation structure for the entire TSP7 series from
the regression in Table A-15. Here, the correlation of the reciprocal of
wind speed with particulate levels is positive, as assumed in most air
quality diffusion models. For the entire S02B and TSP1 series, however,
the correlations with wind speed were found to be near zero in the regres-
sion analysis (see Section VI.A.2.b.(2)).
In general, therefore, an independent analysis of the
time periods in the six representative series during which heating degree-
days and mean wind speeds deviated significantly from climatological ave-
rages confirmed the results from the stepwise multiple linear regression
analysis.
A-100
llUahknl
-------
APPENDIX B
EMISSIONS AND REGULATORY ANALYSIS
UJalden
-------
B-l. PROJECTION AND APPORTIONMENT PROCEDURE
Projection of 1972 fuel use and emissions to the years 1973 and 1974
was accomplished (1) through the application of general indicators of econ-
omic and population change to the base year fuel use inventories to adjust
industrial process fuel use, and (2) by applying estimates of degree-day
differences between the base year and the year of interest to adjust space
heating fuel use. This appendix summarizes the input data and tasks involved
in performing these projections.
1. Point Source Update
Using information obtained from the Massachusetts Bureau of
Air Quality (BAQC), a tabulation was made of all point sources which opened
or closed between 1972 and 1974. This study indicated that no new point sources
opened in 1973. A number of small existing sources which had previously been
grouped within an area source were registered during 1973; however, these
sources did not represent new fuel use and, thus, were not specifically intro-
duced into the inventory.
Most significant of the point source closures between 1972 and 1974
was a bisulfite plant,located several kilometers from downtown Boston, which
ceased operation in 1973. This plant had emitted S02 at an annual average rate
of 0.037 kg/sec (1,286 tons/year) while in operation. Several small incinerators
were also closed during this period as a consequence of the BAQC's particulate
control efforts.
2. Growth Factor Development
Demographic indicators used to estimate changes in fuel use due to
changes in population and the economy were obtained from various state agencies
in the Commonwealth of Massachusetts. Specifically, population growth project-
tions by city and town for the period 1970 to 1975 were received from the Depart-
ment of Environmental Affairs. These factors were summarized by U.S.G.S. topo-
graphic quadrangle based on the relative land area of each city or town in the
quadrangle. During summarization, some municipalities were assigned a zero
B-l
Ulaiden
-------
weighting factor if it appeared that the population growth for that city
would not be indicative of the quadrangle growth. This occurred if, for
example, only a town's conservation land was in the quadrangle in question.
Table B-l summarizes these growth rates.
During the step associated with the application of these factors
to the base year inventory for projection of residential fuel use, it was
noted that differences in growth rates among quadrangles was very slight.
Also, grouping core municipalities (Boston and the 12 surrounding communities)
with neighboring suburbs tended to mask the trend of "urban flight". Therefore,
it was decided to project residential fuel use only on a core-noncore basis, as
also shown in Table B-l.
Industrial and commercial growth were estimated by projections
made by the Department of Manpower Affairs of employment growth for the
Metropolitan Boston SMSA. The annual growth rates derived from these data
were 0.58% for industrial and 3.39% for commercial facilities, respectively.
3. Conservation Analysis
A survey of principal fuel users and distributors was conducted
to determine a quantitative estimate of fuel savings which resulted from
conservation efforts in the AQCR during the winter of 1973-74. The results
of this survey are included as Appendix B-2.
4. Disaggregation of Seasonal Fuel Use
The NEDS point source file contains an estimate of the fraction of
fuel each source uses for space heating. These data were used to disaggregate
the fuel use of each point source into a space heating and a process use com-
ponent. Estimates of the space heating use for each source category of the area
source inventory were based on the average percent spaceheating requirement from
a comparable point source category, or by procedures outlined in Reference 31.
These procedures resulted in an estimate that 80 percent of residential and com-
mercial-institutional fuel is used for space heating. Industrial fuel was
assumed to be used entirely for process applications.
B-2
IWakkni
-------
TABLE B-l
AVERAGE ANNUAL POPULATION GROWTH RATES
SUMMARIZED BY U.S.G.S. TOPOGRAPHIC QUADRANGLE
1970 - 1975
Quadrangle
Bill erica
Blue Hills
Boston North
Boston South
Brockton
Cohassett
Concord
Duxbury
Framingham
Franklin
Gloucester
Hanover
Holliston
Hull
Ipswich
Lexington
Lynn
Mansfield
Marblehead North
Marblehead South
Maynard
Medfield
Nantasket
Natick
Newton
Norwood
Reading
Rockport
Salem
Scituate
Weymouth
Whitman
Wilmington
Wrentham
Average Core Area (13 Cities) Change
Average Noncore Area Change
Percent Population Change
7.90
1.38
0.49
-0.39
1.98
2.05
2.62
4.30
1.28
7.55
0.97
1.88
2.16
-0.20
2.32
1.13
0.16
2.07
1.79
-0.02
0.85
3.67
-1.11
1.65
0.30
1.73
1.60
1.97
1.64
3.89
0.81
1.50
1.88
2.46
-0.30%
1 . 37%
B-3
Maiden
-------
5. Regional Meteorological Analysis
A study of the variation in heating requirements throughout the
AQCR was performed. Table B-2 presents the ratio of annual degree-day totals
to climatological mean degree-days for five meteorological stations located
within or near the Metropolitan Boston AQCR. These data indicate that de-
partures from climatological normals are not uniform throuqh the area. However,
these variations are not extreme, allowing for the assumption that degree-days
totals recorded at Boston's Logan International Airport are representative of
the AQCR.
Table B-3 presents a breakdown of the fraction of degree-days
occurring in each quarter at Logan. These data were used to apportion fuel
use to quarters.
6. Projection/Apportionment
Projection and apportionment of the point and area source in-
ventory was performed by applying growth, conservation, and degree-day factors
according to the relationship:
4
F9 = 0.0025F, z (con,)[(sh)(ddMq,) + (100 - sh)gf]
1=1 ' 1
where:
Fo = Projected fuel usage (mass/year)
F, = base year fuel usage (mass/year)
sh = portion of fuel used for space heating (percent)
dd1 = adjusted ratio of degree-days observed during the pro-
jected period to those observed during the base period
i = quarterly summation indicator
con.. = an estimate of the effect of fuel conservation by quarter
gf = the growth factor; an estimate in the change in activity
of the source based on its SCC code
q. = portion of annual degree-days occurring in each quarter
This procedure was used'to simultaneously create a projected annual
B-4
UJalden
-------
TABLE B-2
COMPARISON OF ANNUAL DEGREE-DAY RATIOS*
FIVE STATIONS IN METROPOLITAN BOSTON
AIR POLLUTION CONTROL DISTRICT
Station
Boston WSO
(Logan International
Worcester
Taunton
Chestnut Hill
Framing ham
1969-1970**
1.023
1.059
1.100
1.130
1.110
1970-1971
1.081
1.053
1.091
1.214
1.112
1971-1972
0.982
0.989
1.053
1.111
1.039
1972-1973
0.980
0.988
1.018
msg
0.998
Mean 1.084 1.110 1.035 0.996+
* Base period: 30-year climatology for each station
** Annual totals are based on year July through June
+ Four-station average
B-5
Maiden
-------
TABLE B-3
PROPORTION OF DEGREE-DAYS OCCURRING IN EACH QUARTER
BOSTON (LOGAN INTERNATIONAL AIRPORT)
1st Quarter 0.516
(Jan., Feb., Mar.)
2nd Quarter 0.134
(April, May, June)
3rd Quarter 0.012
(July, Aug., Sept.)
4th Quarter 0.388
(Oct., Nov., Dec.)
—1
Itat 11
B-6
-------
fuel use inventory plus inventories for each of quarterly periods of interest.
7. Verification of 1973 Projections
Verification of 1973 annual fuel use totals for the AQCR were per-
formed by comparing projected fuel use totals by fuel type and source cate-
gory to similar data estimated by the Bureau of Mines (23). These data were
apportioned to the AQCR from the State total by assuming that the relative
proportions of AQCR to state fuel use remained constant from 1972 to 1973.
Initial results of this study indicated that the degree-day factor
"dd", in the growth equation overestimated the effects of degree-day differences
Indications were found [27] that this factor should not be linear but rather
a logarithmic function of degree-days. A regression performed on independent
fuel use totals indicated that the equation
dd1 = 1.2 log(dd) + 1
fit well the variations in space-heating fuel use in New England.
Table 3-2 in Section III presents projected annual fuel totals
for 1972 through 1974 and a comparison with Bureau of Mines data for 1972 and
1973.
8. Fuel Quality Survey
The extent to which fuel did not conform to SIP regulations during
the winter of 1973-74 and an estimation of the sulfur content of the non-con-
forming fuel was determined by a survey of the major fuel distributors and users
in the area. These data were used to create the actual emissions inventory
for 1974. The results of this survey are contained in Appendix B-3. The
final emissions inventory for each year including the effects of nonconforming
fuel are presented in Table 3-4 of Section III.
lUlalden
-------
B-2. FUEL CONSERVATION
Estimates of the impact of fuel conservation efforts on total fuel
usage for each fuel type in the MBAQCR were compiled by surveying major fuel
users and suppliers in the region. The survey group was chosen to provide
information from a broad cross-section of fuel use categories throughout
the area. When available, published data were also used to aid in quantifying
the conservation efforts.
1. Distillate Oil
The New England Fuel Institute (NEFI) has compiled figures on total
distillate oil use in Massachusetts in calendar years 1972 and 1973 [32].
These data reveal sales of distillate oil dropped 2.3 percent in 1973 from
1972 levels. No 1974 data were available.
In March 1974, NEFI reported [33] that conservation and weather
combined achieved over a 20 percent reduction in distillate oil consumption
in New Englnad last winter. More recently, NEFI has reported [32] that con-
servation measures affected a 14.8 percent reduction and warmer weather a 6.7
percent reduction in distillate oil usage in New England last winter. The
Better Home Heating Council estimated [4] reductions of 10-15 percent in
distillate use in Boston due to conservation measures alone. Northeast
Petroleum reports that distillate oil deliveries to more than 900 retail
dealers they supply in Massachusetts were down 15-17 percent during the 12-
month period starting October, 1973.
From these data, conservation of distillate oil was assumed to
be ten percent of total fuel conumption during the last quarter of 1973 and
15 percent in 1974.
2. Residual Oil
NEFI statistics for residual oil use in Massachusetts indicate
total sales in calendar year 1973 fell 2.1 percent from 1972 levels. NEFI
does not have any information regarding conservation of residual oil last
winter in New England.
—"7
lltidey
-------
The regional director of the Federal Energy Administration (FEA)
in Boston, Mr. Robert Phil pott, has stated [34] that the effectiveness of
conservation efforts has not been studied so far.
An individual effort worth noting was MIT's residual oil usage which dropped
6 percent in 1973 from 1972 levels and fell an additional 23 percent in 1974.
MIT's fuel use in the emissions inventory was adjusted to reflect these
specific data.
Monthly totals of Boston Edison's production of electricity fell
from predicted levels in late 1973 and early 1974. New England Power Co.
did not reply in this area. Electrical power generation sources burn 78 percent
of the residual oil used by point sources in Metropolitan Boston. Assuming
Boston Edison's figures are representative of all electrical power usage in
Metropolitan Boston, fuel conservation efforts accounted for a nine percent
reduction in electrical power fuel use during the fourth quarter of 1973 and
a four percent decrease in 1974. The effect of degree-days on these totals
was assumed to be negligible.
Of all fuel suppliers surveyed, only Northeast Petroleum and Union
Petroleum could provide information on trends in residual oil deliveries last
winter. Northeast reports that residual oil deliveries in Boston were down
10-12 percent during the 12-month period starting October 1973. During
this time period, degree-day totals were about 5 percent below normal. Union
reports residual oil deliveries in Boston were down about 25 percent last
winter. NEFI estimates degree-day totals last winter were about 7 percent below
normal. Assuming these dealers service mostly small source users of residual
oil(i.e., area sources), the degree of conservation attributable to these
sources was an 8 percent reduction in the last quarter of 1973 and 12 percent
in 1974.
3. Natural Gas
Boston Gas officials reported [34J that natural gas usage dropped
10-20 percent in Boston in 1974 from 1973 after correcting for weather effects.
B-9
Maiden
-------
Due to the similarity of user types, the effects of conservation of natural
gas were assumed to be the same as those for distillate oil, i.e., 10 per-
cent at the end of 1973 and 15 percent in 1974.
4. Gasoline
Data from the Massachusetts Department of Corporations and Taxes
indicate that total gasoline consumption in Massachusetts rose 3.1 percent
from 1972 and essentially returned to 1972 levels in 1974.
Maiden]
-------
B-3. REGULATORY AND VARIANCE IMPLEMENTATION
Estimates of the extent to which SIP variances, granted as a result
of the energy shortage, were implemented were made through a survey of major
fuel users and distributors in the MBAQCR. Results of a similar survey
conducted by the Massachusetts' BAQC have been used to supplement the infor-
mation received by Walden, and the BAQC information was assumed to be correct
whenever conflicting information was received from an individual contact.
The following subsections (A&B) summarize the results of the sur-
vey by fuel type. Results are in the form of scaling factors, i.e., the
ratio of the average emission rate of the pollutant for actual and for
full variance implementation to that of the SIP Regulation, assuming con-
stant fuel use.
A. Sulfur Dioxide Scale Factors
1. Distillate Oil
The associated variance relaxed the SIP Regulation statewide
to allow use of 0.5 percent sulfur distillate oil (0.5% S) instead of 0.3
percent sulfur fuel.
For actual variance implementation, the scaling factor of
1.02 was determined from the survey results. An analysis by the BAQC suggests
approximately 9 percent of distillate used was nonconforming at about 0.37% S.
This would also imply use of a scaling factor of 1.02.
For full variance implementation, distillate emissions were
multiplied by 1.67.
2. Residual Oil
The associated variances were only granted to certain sources
and in general allowed use of 1.0% S instead of 0.5% S residual oil in the core
and 2.2% S,instead of 1.0% S residual oil outside the core area.
Residual oil use during the winter of 1973-74 in the Metro-
Uhkkn,
-------
poll tan Boston AQCR can be categorized as follows:
103GAL/YR % OF TOTAL
Point Source Granted Variance 1,078,479 56.2
Point Source No Variance 88,355 4.6
Area Source Granted Variance 127,748 6.7
A ma Source No Variance 623,712 32.5
1,918,294 100.0
a) Point Sources
Table 3-3 in Section III presents the actual and full
variance sulfur contents of residual oil used by point sources that were
granted variances.
b) Area Sources
For actual variance implementation, a scaling factor of
1.13 was determined from survey results for sources in the core, 1.00 for
sources out of the core.
For full variance implementation, residual area sources
emissions were multiplied by 1.18. Although full variance implementation
would be expected to more than double emissions, only 17 percent of residual
area sources were granted variances.
3. Coal
The only variance associated with coal was to allow units
1, 2, and 3 of New England Power Co.'s Salem Harbor Plant (point sources) to
burn 2.5% sulfur and 15% ash coal instead of residual oil after Jan. 23, 1974
For actual variance implementation, coal with an average
sulfur content of 0.77% was burned in units 1, 2, Ind 3. No change in this
factor, to account for oil usage in January, 1974, is necessary due to the
low sulfur content of the coal that was burned. The scale factor for actual
implementation is 1.05.
For full variance implementation, 2.6% sulfur residual oil
-------
could have been used from Jan. 1 through Jan. 22, 1974, and 2.5% sulfur coal
from Jan. 23, 1974, through March 31, 1974. To account for this split, coal
combustion was used in the inventory for units 1,2, and 3, and SOg emissions
were multiplied by 0.94 to account for the 22 days of lower emissions from oil.
The net scale factor was then 3.2.
B. Particulate Scale Factors
1. Distillate Oil
No increase in particulate emissions was assumed to result
from full or actual variance implementation.
2. Residual Oil
No increase in particulate emissions was assumed to result
from full or actual variance implementation.
3. Coal
The only variance affecting coal consumption during the energy
shortage involved 15.6 percent ash coal burned in units 1, 2, and 3 of NEPCO's
Salem Harbor Plant. The full and actual variance implementation cases for this
facility are identical. The particulate emission rate for 15.6 percent ash
coal is about 150 times greater than that for an equivalent quantity of oil.
However, electrostatic precipitators used on these units resulted in a 92 percent
reduction in particulate emissions. After adjusting for oil use during the first
three weeks in January, plus the control efficiency of the precipitators, the net
scale factor was 9.11.
B-13
Illhldem
-------
APPENDIX C
MODELING ANALYSIS
Illhlden/
-------
C-l. AIR QUALITY MODEL AND SIMULATION SCALING PROCEDURE
A. Dispersion Model
References 28, 29, and 30 describe the basic Air Quality Display
Model*, which was used in this analysis to predict ground level concentrations
of TSP and S02 throughout the Metropolitan Boston AQCR. Consequently, an extend-
ed discussion of the model is not repeated here. However, several modifi-
cations to the basic form of the model were introduced in the analysis. These
included:
• The Briggs plume rise equations were substituted for the
Holland equation specified in the original model.
• A source contribution file was created to facilitate strategy
simulations and to bypass the more costly control strategy approach of the
IPP model.
The input data to the model included the STAR meteorological wind-
stability data for each of the first quarters of 1972-1974 as recorded at
Bsoton's Logan International Airport and the emissions inventories described
in Section III.
The half-life of sulfur dioxide was assumed to be 0.5 hours in
this study. This value, which is considerably less than the generally
accepted value of three hours, was suggested by the Massachusetts Bureau of
Air Quality Control (BAQC). This value of the half-life of S02 provided
a better correlation with observed data in modeling runs of Metropolitan
Boston performed by BAQC.
The effective stack height of each individual area source grid
was computed by averaging the mean effective stack height of each of the
following emissions catenaries, w^i^hted by its relative emission strength
within each grid.
As incorporated in IPP.
C-l
Maiden
-------
Mean Effective
Source Category Stack Height (m)
Residential 15
Commercial-Institutional 47
Industrial 58
Incineration 23
Roadway 4
The mean effective stack heights were computed by the BAQC using average
source parameters for each category and the Briggs' plume rise equations.
B. Scaling Procedure
The simulation of the no variance and full variance cases was per-
formed by developing scale factors to reflect the change in emissions re-
sulting from the various degrees of variance implementation. These scale
factors were then multiplied by the corresponding source contribution
and summed for each receptor to determine the change in concentration resulting
from the simulation.
The basic scaling equation used was
N
St — V
XT I^'^T* "1 \f
i k i=l J J
IN J I
where
X*.. = total contribution of pollutant k from all sources at
lk receptor i under the strategy simulation
X- -i, = contribution of pollutant k from source j to receptor i
1J K
for 1974 actual conditions
a.., = ratio of new emission estimate to original emission
J estimate for pollutant k and source j
N = total number of sources
The X--L, are contained on the source contribution file tape, and the aik are
the scale fac
Appendix B-3.
(. ., are contained on tne source contrioution riie tape, ana tne a-
the scale factors for the no variance and full variance cases presented in
UUbL
-------
C-2. RECONCILIATION, VALIDATION, AND CALIBRATION
In order to calibrate the predictions of the air quality model, meas-
urements of S02 and TSP concentrations recorded within the Metropolitan Bos-
ton AQCR were obtained from EPA and the Massachusetts Bureau of Air Quality
Control (BAQC). The initial group of sampling receptors were carefully chosen
to provide (1) a common measurement procedure, for SOg and TSP, i.e., the West-
Gaeke bubble technique for measurement of S02 and the high volume sampler
for measurement of TSP, and (2) a sufficient historical record, i.e., contin-
uous operation from January 1972 through March 1974.
1. Sulfur Dioxide
a. Initial Reconciliation
Thirteen monitoring sites, listed in Table C-l, were initially
selected for further study. Consultation with field personnel of the MBAQC
indicated that the Government Center site was not to be considered represent-
ative of air quality in that area due to changes in air flow patterns induced
by new multistory office construction in the area between 1972 and the present;
all other sites were initially assumed to be representative of ambient air
quality in the vicinity of the receptor.
Air quality monitoring using the bubbler method is usually
performed for 24 hours once every six days, resulting in an average sampling
frequency of five observations per month. Criteria for the representativeness
of quarterly data were thus developed to exclude a quarterly (three-month)
average if it was composed of less than three observations (or 60%) in any of
the component months. Quarterly averages were computed by averaging the three
monthly means, thus, preventing a bias caused by unequal numbers of samples
among the months. A number of sites were excluded by the "three observation"
criteria. These sites are annotated in Table C-l. Further study of the represent-
ativeness of monitoring sites indicated that two receptors, those in Quincy
and Revere, were located on the roof of buildings only a few meters from the
chimney of residual-oil burning boilers. These sites were also excluded from
the validation process.
C-3
/Utiden
1
-------
TABLE C-l
CONCENTRATIONS OF SULFUR DIOXIDE (S02) AND TOTAL SUSPENDED PARTICULATES
MEASURED IN THE METROPOLITAN BOSTON AQCR (FIRST QUARTER)
TSP*
Site
Government Center**
Kenmore Square**
South Bay
East Boston
Brookline
Medford
Need ham
Norwood**
Waltham
Woburn
Marblehead
Quincy**
Revere**
1972
--
104
39
39
34
24
13
17
32
17
--
50
58
so2*
1973
66
73
68+
22
51
13+
18
34
20
—
--
58
42
1974
33
30
27
15
22
33
19
34
20
31
17
36
43
1972 1973 1974
161 147 84
48 50 32
41 39 29
42 67 42
32 45 29
55 55 56
54
* units: yg/m3
** excluded from validation due to localized source influence
+ excluded from validation due to insufficient data (see text)
no data recorded
c-4 U/aldenj
-------
An analysis of the measured S02 data was also conducted to
determine the possible deviation of the true mean of S02 levels at a given
site from the measured mean. The number of samples taken during a given
quarterly period at a sampling station is usually less than 12. This may
allow a large discrepancy to appear between measured and actual means. An
analysis based on the observed standard deviation at each site indicated that
at the 5 percent significance level, the true mean often could vary from zero
to twice the reported mean. This indicates that some of the apparent error
in the model predictions results from sampling error. Kenmore Square is an
important exception to this observation. At this site, there were sufficient
observations to indicate that the reported mean would not be significantly
different from the true mean (although sampling at Kenmore was never con-
ducted on Sunday). This implies that any difference between predicted and
observed values at Kenmore are due to other factors, e.g., the inventory,
the model, or micro-meteorological differences.
b. 1972 Validation Run
The 1972 first quarter emissions inventory was processed
to predict S02 concentrations at the reconciled monitoring sites for the
validation (see Table C-l). The initial run indicated that S02 concentrations
were overpredicted by a factor of two and the correlation coefficient was about
0.7.
Analysis of the source contribution file for sulfur dioxide
indicated that area sources relatively far from the receptor (15 km) were
contributing more sulfur dioxide to the receptors than would be reasonably
expected. Consultation with the Massachusetts Bureau of Air Quality Control
(BAQC) indicated that they had experienced similar problems in modeling the
Metropolitan Boston area as part of the Air Quality Maintenance Plan process.
The Bureau indicated that reducing the half-life of S02 in the atmosphere
from three hours to 0.5 hours resulted an a better correlation with observed
data [35].
These initial validation steps also indicated that two sites,
C-5
/UJalden,
-------
Kenmore Square and Norwood, were significantly underpredicted. This same
result has been obtained in other attempts to model the Metropolitan Boston
area [35] and has been attributed to a combination of sampling error, micro-
meteorology, and modeling error; these sites were, therefore, excluded from
the validation step.
A second validation run using the 1972 first quarter inven-
tory and the above changes resulted in a statistically significant linear
regression correlation coefficient of 0.86 and a significant improvement in
the degree of overprediction. A regression analysis of the measured (Y)
versus predicted (X) concentrations for the seven reconciled sites yielded
the following line of best fit:
Y = 0.60X + 10.
However, since the natural background for S02 is approximately zero, and the
predicted intercept is only 10, the forced zero-intercept form of regression
analysis was used. The results of this analysis indicated the line of best fit:
Y = 0.86X
with a correlation coefficient of 0.75, which is also statistically significant,
c. 1973, 1974 Validation Runs
The 1973 and 1974 first quarter inventories were each pro-
cessed to provide predictions of air quality at the reconciled monitoring
sites in Table C-l. Regression analyses of these predictions versus measured
data indicated that the 1973 line of best fit was statistically the same as
that for 1972. However, the correlation coefficient for this year was 0.54
using the slope-intercept method, and 0.48 using the slope-forced-zero-inter-
cept procedure. Neither of these values is statistically significant, con-
sidering the size of the data sample analyzed.
In 1974, S02 levels at many of the air quality receptors in
the MBAQCR v/ere below the minimum sensitivity level of the monitoring instru-
ment (25 yg/m3). Thus, differences among measured means may not represent
UJaldffi.
-------
true variations. An analysis of the 1974 first quarter measured data Indi-
cates extreme scatter in these values.
The expected spatial distribution of S02 concentrations where
the mean 1s greater 1n urban areas and lower in suburban areas is not apparent
in the 1974 data. Two urban sites, East Boston and South Bay, located only
five kilometers apart, reported concentrations of 15 ug/m3 and 27 ug/m3,
respectively, in 1974, whereas each site had reported 39 yg/ma in 1972. Ken-
more Square, an urban site, averaged 104 ug/m3 in 1972; Norwood, a suburban
site, averaged 17 ug/m3. In 1974, Norwood was actually higher than Kenmore
Square, 37 ug/m3 versus 30 ug/m3.
This extreme variance in spatial and temporal patterns of
measured data has been inexplicable. Thus, an extremely poor correlation
was obtained when attempting to validate the 1974 measured versus predicted
concentrations. However, a visual study of the data indicates that the 1972
line of best fit provides a reasonable fit of the 1974 data.
Figure C-l presents the results of the validation for each
year and compares these observations to the 1972 line of best fit, I.e., the
forced-zero intercept, which was used for calibration.
2. Total Suspended Particulates (TSP)
Seven monitoring sites were initially selected for further
study. Table C-l lists these sites. Criteria for excluding TSP monitor-
ing data were identical to those used for SO^ (see preceding subsection).
Exclusion of some sites in an initial reconcilation and of
two others as a result of this validation analysis (see preceding sub-
section) resulted in only three usable monitoring sites, which is an in-
adequate basis for performing a meaningful regression analysis. However,
due to the favorable agreement among the SOg regression parameters ob-
tained in each of the years, it was decided to perform a validation by
combining data from all three years for the three sites.
C-7
-------
Observed Concentrations (yg/m3)
no
o
co
O
(Jl
O
en
o
o
-$
n>
ex
_^.
o
rt
n>
Q-
o
o
3
n
-J
DJ
ct-
O
3
in
ro
o
co
o
en
o
-------
This regression provided a best line of fit of
Y = 1.07 X + 20
and a statistically significant correlation coefficient of 0.73.
This equation was used for the TSP calibration procedures,
The forced-zero-intercept regression was not used, due to
the existence of a significant natural TSP background.
Wakkn,
-------
Observed Concentrations (yg/ma)
o
I
ro
CO
CO
-o
m
oo
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-450/3-75-068
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Impact of
Energy Shortage on Ambient Sulfur Dioxide and Parti -
culate Levels inMetropolitan Boston AQCR
5. REPORT DATE
July, 1975
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Dr. Richard D. Siege!; Mr. Peter H. Guldberg; Mr. Ken-
neth W. Wiltsee, Jr.; Dr. Ralph B. D'Agostino, Boston U.
8. PERFORMING ORGANIZATION REPORT NO
C-597
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Walden Research Division of Abcor, Inc.
201 Vassar Street
Cambridge, Massachusetts 02139
10. PROGRAM ELEMENT NO.
2AC129
11. CONTRACT/GRANT NO.
68-02-1830
12. SPONSORING AGENCY NAME AND ADDRESS
;NCY NAME AND ADDRESS
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
is.SUPPLEMENTARY NOTES Presented ^n part at tne 68th AIChE Annual Meeting, Los Angeles,
California, November 1975, and at the AMS 4th Conference on Probability and Statistics
in Atmospheric Science. Tallahassee. Florida. November 1975.
16. ABSTRACT
The purpose of this project was to evaluate the impact of the energy shortage on
ambient sulfur dioxide (S02J and total suspended particulate (TSP) concentrations in
a major urban area, Metropolitan Boston. A combined approach based on a statistical
analysis of measured air quality data, regulatory and emissions analysis, and diffu-
sion modeling of changes in ambient pollutant concentrations was used to attain this
objective.
The objective of the air quality analysis was to identify and interpret changes
in trends of measured S02 and TSP levels between January 1970 and March 1974 with
regard to SIP regulations, variances granted because of the energy shortage, and
meteorological conditions. The objective of the regulatory and emissions analysis was
to develop quarterly emissions inventories for S02 and particulates in Metropolitan
Boston for the period January 1973 to June 1974 to support the subsequent modeling ana-
lysis task. The objective of the diffusion modeling task was to obtain a clear under-
standing of the impact and potential impact of the energy shortage on air quality
levels by isolating the relative effects of factors such as meteorology, growth and
conservation, and variances on ambient S02 and TSP concentrations.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
13. DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
212
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
-------
INSTRUCTIONS
1. REPORT NUMBER
Insert the EPA report number as it appears on the cover of the publication.
2. LEAVE BLANK
3. RECIPIENTS ACCESSION NUMBER
Reserved for use by each report recipient.
4. T|TLE AND SUBTITLE
Title should indicate clearly and briefly the subject coverage of the report, and be displayed prominently. Set subtitle, if used, in smaller
type or otherwise subordinate it to main title. When a report is prepared in more than one volume, repeat the primary title, add volume
number and include subtitle for the specific title.
5. REPORT DATE
Each report shall carry a date indicating at least month and year. Indicate the basis on which it was selected (e.g., date of issue, date of
approval, date of preparation, etc.).
6. PERFORMING ORGANIZATION CODE
Leave blank.
7. AUTHOR(S)
Give name(s) in conventional order (John R. Doe, J. Robert Doe, etc.). List author's affiliation if it differs from the performing organi-
zation.
8. PERFORMING ORGANIZATION REPORT NUMBER
Insert if performing organization wishes to assign this number.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Give name, street, city, state, and ZIP code. List no more than two levels of an organizational hirearchy.
10. PROGRAM ELEMENT NUMBER
Use the program element number under which the report was prepared. Subordinate numbers may be included in parentheses.
11. CONTRACT/GRANT NUMBER
Insert contract or grant number under which report was prepared.
12. SPONSORING AGENCY NAME AND ADDRESS
Include ZIP code.
13. TYPE OF REPORT AND PERIOD COVERED
Indicate interim final, etc., and if applicable, dates covered.
14. SPONSORING AGENCY CODE
Leave blank.
15. SUPPLEMENTARY NOTES
Enter information not included elsewhere but useful, such as: Prepared in cooperation with, Translation of, Presented at conference of,
To be published in. Supersedes, Supplements, etc.
16. ABSTRACT
Include a brief (200 words or less) factual summary of the most significant information contained in the report. If the report contains a
significant bibliography or literature survey, mention it here.
17. KEY WORDS AND DOCUMENT ANALYSIS
(a) DESCRIPTORS - Select from the Thesaurus of Engineering and Scientific Terms the proper authorized terms that identify the major
concept of the research and are sufficiently specific and precise to be used as index entries for cataloging.
(b) IDENTIFIERS AND OPEN-ENDED TERMS - Use identifiers for project names, code names, equipment designators, etc. Use open-
ended terms written in descriptor form for those subjects for which no descriptor exists.
(c) COS ATI FIELD GROUP - Field and group assignments are to be taken from the 1965 COSATI Subject Category List. Since the ma-
jority of documents are multidisciplinary in nature, the Primary Field/Group assignment(s) will be specific discipline, area of human
endeavor, or type of physical object. The application(s) will be cross-referenced with secondary Field/Group assignments that will follow
the primary posting(s).
18. DISTRIBUTION STATEMENT
Denote releasability to the public or limitation for reasons other than security for example "Release Unlimited." Cite any availability to
the public, with address and price.
19. & 20. SECURITY CLASSIFICATION
DO NOT submit classified reports to the National Technical Information service.
21. NUMBER OF PAGES
Insert the total number of pages, including this one and unnumbered pages, but exclude distribution list, if any.
22. PRICE
Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
EPA Form 2220-1 (9-73) (Reverse)
------- |