-------
            - GO <— LOCOCOliirOOCMr— Oi— CO f-- <*£> "zf < "> f^~> CT> CT>
            -- co r-r ro CM n LT> n oo M co cr> i— 01 vo oo r--. co n •* "*o
            o-^j-r-i— i—              i— en i— i— CM r""> UD        i.n cr> «=r
             •v  *•  *»  **
           CO CM r— P	
                                               CM
                                                                      en
          o
                 CM o CM > — i —
                                              CM
                                         cooooi— r-^r^or-^i
                                         CO [-^ O 00 i— CO i— *£>!
                                                 r,            n   rt
                                               i—           10 r^. <
  i— LO to cr> vo r- un «3-  UD r— cM^cocor^oouD
 CMCMCMi— i— CM'— O <— i— CM CM  CM CM CM CM i— i— CM CM '—

 ooooooooooooooooooooo
                                          i — oo co en
                                                                   i
                                       co i — r--i — CMOOCM    i —
                                                                      - CM
                                                                                               CM
                                                                                               r^
                                                                                               en
                                                                                               JD
                                                                                               o
                                                                                               o
                                                                                               o
                                                                                       CU
                                                                                      re

                                                                                      •I—
                                                                                      o

                                                                                      -a
                                                                                       c
                                                                                       ra
 o  o
-^ -i^
 rO  ra
a o.
         4J 4->
         S-  3
         o  o
         "
cu
CD
CO rO
co £
CU ra
 a.
 a_
•r-  I/)
 to  ra
 i/>  co •
       •M  E J3 CO ra
 ra
 E
 o
-E
 ra
       O)
           CU r—
          I— <
                                                    O
                                           "3     DTO
                                           c     c: ra
                         s: < _j o H-
                 E ra O r—  ra > __  .
                 o-a>,o4->cuiDi-
                 s:i-isorjs2rcc
    O     E
    o     o
   •I-    +J
    X ra  CD
 «  OJ E  E  E
•as: o -r-  o
 "3     N X:  C7)
           to  CU
           ra  S_
          3 O
                                                                           ra
                                       •rO
                                       o
                                                                           tt)
                                                                           o
                                                                           i-
                                                                           3
                                                                           o
                                                                           oo
                                                              A-6

-------
          Average degree-days from the  1970 Statistical
           Abstract, of the U.S.
          Per capita income from the 1971  Statistical
           Abstract of the U.S.
          Price of residential  gas from the 1971  Gas
           Facts, American Gas  Association
          Average rooms/housing unit from the 1970 Census
           of Housing, General  Characteristics

          In the stepwise regression, the variables were selected in
the following sequence:  degree-days (x,), price  (x,-,),  per capita income
(x.J, average rooms/housing unit (x»).   The final  equation was:

     y = 0.00011 x] - 0.34450 x£ + 0.00021 x3 + 0.22904 X4 - 0.93373

where  y is expressed in thousand therms/homeheating  customer
      X-, is expressed in degree-days
      x~ is expressed in dollars per thousand cubic feet (or million Btu)
      x3 is expressed in dollars
      x. is expressed in average number of rooms
                                                         2
          The adjusted multiple correlation coefficient R  equalled 0.672.
The F-statistic, F(4.45), was 25.377, which indicates a 0.995 confidence
level.  The adjusted standard error of estimate was 0.205 and the range
of residuals was 0.893.  The t-statistics for the coefficients showed a
0.9995 confidence level for the coefficients of x,, x2, and x3 and a 0.975
confidence  level for the coefficient of x..

          Leaving out the variable indicating average number of rooms per
                           2
housing unit we  found the R  reduced to 0.646 and the adjusted standard
error slightly higher, 0.211.  The confidence level remained more or less
the same.   Interestingly, the range of residuals was reduced to 0.860,
although the standard error increased.  This reduction is not significant
enough to warrant deletion of this fourth variable.

          The  correlation matrix  shown in Table A-3 showed no significant
intercorrelation of the  independent variables.
                                  A-7

-------
                              TABLE  A-3

                          CORRELATION  MATRIX

y
xi
X2
X3
X4
y
1.000
0.672
-0.193
0.362
0.214
xl
0.672
1.000
0.196
0.217
0.286
X,
-0.193
0.196
1.000
0.281
0.276
X3
0.362
0.217
0.281
1.000
-0.087
X4
0.214
0.286
0.276
-0.087
1.000
                              TABLE A-4

             RESIDENTIAL DISTILLATE OIL CONSUMPTION BASED
                   ON MULTIPLE REGRESSION EQUATION
     State                           Thou Barrels Consumed (1970)


Massachusetts                                  14,896

Maryland                                        7,048

Missouri                                        1,234

Washington                                      4,^786

Maine                                           3,951

Connecticut                                    10,475
                                  A-8

-------
          Using the resulting equation  and tank wagon prices for #2 fuel

oil shown in the 1970 November issue of Fuel  Oi 1 and Oil  Heat for selected

cities and assuming that these prices are representative  for the entire

state in which the cities are located,  we obtain the fuel oil figures

shown in Table A-4.


      d.  METHOD BASED ON HEAT LOSS AND DESIGN RANGE


          The Yankee Oilman, March 1973 Fuel  Trades Fact  Book shows the

following formula:

Fuel Units Consumed Annually = Heat Loss x Annual Degree  Days, x Use Factor
                                    Btu per Fuel Unit x Design Range


where the heat loss is the hourly amount of Btu's lost per square feet of

area multiplied by the total square feet; the use factors are fuel and use
dependent and listed in a table in the publication; the design range is de-

fined as the difference between inside temperature (70°F) and the design

outside temperature for the area.


          Wai den obtained the state-by-state data needed to use this

formula from the following sources:


          (1)  Average square feet per housing  unit
               for selected states (needed for  heat
               loss calculations) - Independent Gas
               Association of America, Comparison
               of Seasonal Househeating^ Costs,
               December 1972.

          (2)  Design range - "Systematic Study of
               Air Pollution from Intermediate  Size
               Fossil Fuel Combustion Equipment,"
               Walden Research Corp., Cambridge,
               Mass.

          (3)  Degree days - 1970 Statistical Ab-
               stract of the United States, U.S.
               Dept. of Commerce, Washington,  D.C.


Other factors were taken from the Yankee Oilman publication  in which the
              »
formula was  shown.  The estimated distillate oil used for spaceheating  is

listed  in Table A-5 for selected states.
                                   A-9

-------
                     TABLE  A-5

ESTIMATES OF DISTILLATE OIL CONSUMED FOR HOME  HEATING
      BASED ON A PER HOUSING UNIT REQUIREMENT
             ACCORDING TO AVERAGE FT2
State
Pennsylvania
Indiana
Illinois
Michigan
Iowa
Missouri
Nebraska
Georgia
Florida
Kentucky
Tennessee
Alabama
Oklahoma
Idaho
Arizona
Washington
Distillate Oil
(thou barrels)
21,487
7,055
7,165
11,962
2,839
3,236
833
1,123
2,466
1,683
1,143
143
21
1,808
33
11,062
                         A-10

-------
         Table A-6  shows a  summary of the four methods.
                             TABLE A-6
         SUMMARY OF ESTIMATION METHODS  FOR RESIDENTIAL FUEL
                      OIL USED FOR SPACEHEATING
                              Based on:

Massachusetts
Maryland
Missouri
Washington
Maine
Connecticut
Florida
Alabama
EPA*"
25,678
9,853
2,494
9,533
8,700
15,984
1,927
128
to)
Equipment^ '
42,005
10,307
3,466
16,109
8,326
19,665
1,389
245
Multiple/o}
Regression^ '
14,896
7,048
1,234
4,786
3,951
10,475
n.a.
n.a.
(4\
Square Feetv '
n.a.
n.a.
3,236
11,062
n.a.
n.a.
2,466
143
(1)   EPA Guide for Compiling a Comprehensive Emission Inventory
(2)   Fuel Oil  and Oil  Heat;  Intermediate Boiler Study,  Wai den
(3)   Walden Research Corporation
(4)   Independent Gas Association of America; Yankee  Oilman
                                 A-11

-------
     2.   County Fuel  Oil

         Walden collected sufficient  community-by-community data to perform
stepwise regression analyses of the  type  shown  below:

       Residential  gas use = a  degree days  + b median  income
                              X                A
                             + c  average rooms per  housing unit + d
                                X
The regressions were performed on a  community-by-community basis for  two
reasons:  (1) the majority of gas companies  do  not  individually service
enough counties to provide a significant  sample for  a  regression analysis,
(2) it is quite common for a utility to service only a portion  of  a  given
county,  in which case that utility's sales in  the  county would  not  reflect
total consumption.

         Walden contacted various gas companies across the  country to obtain
community residential gas sales figures for  the year 1970.  Table  A-7 shows
the companies which provided data for the regressions.  Detailed  degree-day
data were obtained from the Climatological Data Center in Asheville, North
Carolina (6).

                               TABLE A-7
           GAS COMPANIES WHICH PROVIDED RESIDENTIAL GAS DATA
               Utility
      Area Serviced
     Wisconsin Gas Co.
     Northern Illinois Gas Co.
     Boston Gas Co.
     Pacific Gas and Electric Co,
     Southern Union Gas Co.
     East  Ohio Gas Co.
Southern Wisconsin
Northern Illinois
Metropolitan Boston
Central California
Texas, New Mexico, Arizona
Metropolitan Cleveland
Median  income  figures for all communities with populations over 2500 were
obtained  from  the Census of  Population  (7).  Average number of rooms per
dwelling  unit  was obtained in the 1970  Census of Housing (8) for all com-
munities  over  2500.
                                 A-12

-------
         The regression results  are summarized  in  Table  A-8  and  discussed
in detail  below:

         In Wisconsin,  where 21  communities  were analyzed, the first  vari-
able selected in  the stepwise regression  analysis  v/as  average number  of  rooms
                    2
per dwelling unit (R  - .581).  The second was  annual  degree-days.  When this
was entered into  the regression, the multiple correlation coefficient in-
creased to .687.   There was a wider range in the size  of dwelling units  than
there was in the  number of degree-days for the  municipalities considered.
In using degree-days as a variable, it is important to note  that the  degree-
day figures are taken from the nearest weather  station which may be as much
as thirty miles away.  Degree-days also do not  account for a wind chill  fac-
tor, which can be very significant in many cases.

         When gas use for residential space  heating only was used as  the de-
pendent variable in these same 12 communities,  the order of  significance for
the independent variables was not changed, but  the correlation  coefficients
in each case were increased slightly.

         Southern Union Gas Co.  provided Walden with data for 22 communities
in Arizona, New Mexico, and Texas.  The most significant variable in  this
                       2
case was degree-days (R  =  .755).  For this  area,  the degree-day variation
was very large, ranging from 1558 to 7076 degree-days.  The  second  most
significant variable, average number of rooms,  varied by only  .7 rooms,
from smallest to largest.

         The third analysis was performed for 23 communities in the Cleve-
land, Ohio area.  Because these communities  were all within  thirty  miles of
one another, the range of degree-days was almost negligible. There were
only three stations  in the  area from which degree-day readings  could  be
used, further reducing the  significance of the degree-day  data.   Median  in-
come was selected as the most significant independent variable.

         California  data for 87 communities  were obtained.   Median  income
explained  a major proportion of the variability in fuel use  for this  analy-
sis.  The most striking characteristic about these communities  was  that  the
degree-day figures did not  exceed 3200, and, in most cases,  they ranged
                                 A-13

-------
co
 i
UJ
CO


U~l

*J
cc

ro
^tfc

O)
[Q
•i —
s~
ra
^"



CM
-It:

OJ
J2
ro
*f™
S-
ro
^>






	
=*fe

OJ
o
ra
•1 —
i-
ro



-!->
a.
OJ
o

a)
"j2
i — t

01
4_J
rc
^y




c^-
tn
t^.
.
*^j-
CM
CM


LO
CM
to
in
CM

CO
in






























f^.
ro
CM
CO


to
E
O
0


r--
ro
CM

CO
1




1
1 o
to
3










ro
to
cn
•*
C\J
CM
vj-
to












_
^^
CM
CM
O
^

CO
>^
rO
Q

.
Q




to
sl-
i—
•
in

to'
E
0
O
r*"

CM

to

ro
CM
1





CJ
to
2










in
r^
r--
•^j-
CM
^
CO
to

Ln
CM
1 —
CD
^'
Oi
^
o
0
c
t— 1

^
cn
cn
ro
O
— !,

1.1
>^
ra
Q

.
Q




^-^
CM
CM
CO
1-^

to
E
0
o

cn
to
cn

cn
CM
i





o
to
•3.










LT)
0
CM
«;}-
CM
cn
r—
VO






























cn
in
«^j-
•
^j-

to

6
o


1 —
ro
CM

to
1




*
0
co
^










ro
Ln
ro
, —
CM

ro
f^













ro
j~>.
CM
O
— I.

to
>^
ra
Q

.
a




cn
i —
CO
in
in

CO
E
O
O

in
^j-
r—
.
CD
CM
1




*
0
to
3










CO
1 —
cn
, —
CM
0
in
f~-
CM
CM
cn
CD
0
•~-^
ai
E
o
o
cr

CM
O
O
•=!•

vj.

CO
>^
>
ro
Q

•



CM
CO
ro
,_!
to

£^
O
•1 —
c
~D

CO










cn
•-•d-
*=f
r_
^
CT>
cn














to
ro
cn

r-^
CM
v__>

to
E
O
o
a:




, —
r— •
O
•

>>
rO
a

•

r—
o

CM
•
f^
'

C
o
•1 —

'~>
•
o

CO
CD
CO
.
CO
1 —
CM


CO
CM

CO
"~
to

to






























CO
cn
CD
CD

CLI
E
0
CJ


r^^
f-^
cn
cn
ro
cn
co





o
c~
0

o
cn

.
Ln
o
^



f^^
CM
°?
cn
to

to






























in
Ln
o
CD

a>

o
u


to
CM
in
to
in
, 	
"*




n 	

'i-o
"
A-l










CM
0
"^
cn
^
, —
r~-












«d-

cb
o
o
— I.

I/)
>^
ra
Q


Q




t »
co
in
0
o

CJ

o
CJ
c

CM
, —
LO
«^J-
r~~-
r-^





^j

•~
o
4










to
CD
"~|
cn
oo
ro


to
CO
ro
i —
r-
'
LI
g=
O
o

•^
cn
r---
O
O
— 's

0")
^^
ra
c — ^

.
Q




co
«^-
o
0

OJ

o

c

r^^
ro
ro
•s^-
to
CO
1




^J

'o
CJ

^J-
CM
ro
.
cn

^


to
en
CO
CM
r™
CD

to






























S-
in
o
o

cu

o
o


o
to
ro
r\
to
«3-
^







t— H










CM
CM
cn
txi
1 —
CM

to












CM
to
to
r^
co
«=}-
i
x^_^

CO
E
0
o
Qi




CM
to
o
o

ai

Q
O


CM
in
«^-
o
1 —
cn
co







~

tn
cn
0
.
co
t~~
CM


cn
O
r--
O
"*
O
CM
CO





























co
cn
ro
r~—
•
CM
	
I/)
E
0
o

to

i^O
r-.
i —
^j.
"*"






to
rO


an
r^J-
o
.
LO
^i"
"


r— •
CM
CM
to
ro
o

in




























—
fvt
r^
i —
o
•

CO
ro
a

^



CM
LT>
co
, —
to
CO
a.
^
0
S-

1 —
5










*^J-
^J~
Ln
cn
CM
«^-
r^
to













*~~*
to
CM
CO
CO
CO
^— . f'

CO
E
o
o
a:


—

ro
i —
O
•

CO
ro
Q

^


CM

r-~
tn
cn
i

a.
~^
o
S-
CI3

1 —
^

CM
O
in

CO
ro
r— •


CO cn
«tf- CO
CM O
CM r-~
ro CM
^
CM CO
in to












^
cn
r~-
o

o
ro
^-— -^

I/)
E
O
O
a;



P**- CO
iri oo
, — , — .
0 0
•

to to
ra ra
a a

Q a


i^^
<* CO
CO LD
CM •
in r>«
to i
CO 
-------
from 1500 to 2500 degree-days.   This  meant  that  a  much  higher  proportion  of
total  gas use in these communities  was  devoted  to  non-space-heating  applica-
tions  than in any of the previous areas analyzed.   Degree-days,  in general,
do not explain the variability in residential gas  use in  warmer  climates.
                                                             2
Mean income was the most significant  variable  in this case (R   = .676).
The correlation coefficient between degree-days  and gas use was  only .022.

         Data for 68 communities in Northern Illinois were obtained  from  the
Northern Illinois Gas Co.  More than  three  quarters of these ccrrmunities
were in the immediate Chicago area.  This limited  the degree-day figures  to
only five weather stations in the area  with a  total range of difference  of
less than 500 degree-days.  In compiling the degree-day data in this case,
it was exceedingly difficult to assign  a certain weather station's degree-
day figure to a given community with any assurance of accuracy.   Secondly,
the Chicago area is generally subjected to  strong  winds coming off  Lake
Michigan during the normal heating season.   This is no doubt a significant
factor in gas consumption for space heating  purposes, which our degree-day
figures did not reflect.  Median income was the most significant variable
in this case  (R2 =  .610).

         The  final group studied included 19 communities in the Boston area.
The most significant variable in this case was average rooms per dwelling
unit with a correlation coefficient of  .320.  In the Boston area, unlike
the other areas studied, gas is  not the primary heating fuel.   Furthermore,
the proportion of older dwelling units  in the Boston area is probably much
higher than  in the  other areas  studied.  These older units probably have
spaceheating  characteristics which differ significantly from those in
newer housing units, due to boiler conversions and consequent diminished
combustion  efficiency.

         A  regression was  performed on  all  six groups combined.  This gave
a total  of  231 communities.  As  might  be expected, the degree-days  repre-
sented the  most  significant variable because of the wide range of readings
available on  a  nationwide  basis  from below 1500 to over 8000 degree-days.
The multiple  correlation coefficient produced at this first step was .510.
With  the addition  of the average rooms  per dwelling unit, the correlation
                                 A-15
-------
coefficient was markedly improved to .674.   Median  income  was  not introduced
into the regression because as the correlation matrix (Table A-9) shows,
there is a very high correlation between average rooms per dwelling unit  and
median income (r = .806) indicating that median income is  not an independent
variable.

                                 TABLE A-9
                CORRELATION MATRIX FOR SIX GROUPS COMBINED

Gas Use
Degree-Days
Median Income
Rooms Per D.U.
Gas Use
1.000
.717
.572
.637
Degree-Days
.717
1.000
.329
.361
Median Income
5.72
.329
1.000
.806
Rooms Per D.U.
.637
.361
.806
1.000
         A regression was also performed on all the groups excluding the
Metropolitan Boston communities.  This left a total of 212 communities  in
the regression.  The results, however, were not as good as those for all
six groups.  The first-chosen variable, which again was degree-days, gave
a correlation  coefficient of  .527.  When the next variable, average roon,'s
per dwelling unit, was entered into the regression, the multiple correla-
tion  coefficient increased  to only  .666, lower than for all 231 communi-
ties, with both variables entered into the  regression.

         Using the formula:

               residential Dist. Oil (gallons) = [(.01288 x degree-days  +
                               30.41 x avg. rooms per housing unit - 79.54)7
                               .14] x # of  housing units

Walden  calculated  residential distillate oil use for all counties in Maine
(see  Table A-10) and compared the sum of these figures with the Maine state
total obtained by  using  the EPA method (see Table A-l, column C).  The  figures
were  within  12% of each  other.  Similar calculations were performed for
Massachusetts, where the  two state totals differed by less than 1%.  These cal-
culations  increased our  confidence  in the selected method to be used to cal-
culate  residential distillate oil by county.
                                      A-l 6
-------
                 TABLE A-10

RESIDENTIAL DISTILLATE OIL BY COUNTY - MAINE
               (thou gallons)
County
Androscoggin
Aroostook
Cumberland
Franklin
Hancock
Kennebec
Knox
Lincoln
Oxford
Penobscot
Piscataquis
Sagadahoc
Somerset
Waldo
Washington
York
Total Maine
Total Maine (EPA Method)
Distillate Oil
29,203
30,851
64,903
9,365
13,853
34,706
12,255
8,285
15,946
44,396
6,788
9,189
16,591
8,565
11,851
38,910
355,657
398,622
                      A-17
-------
                        REFERENCES  - APPENDIX A
1.  1970 Census of Housing  -  Detailed  Characteristics, U.S. Dept of Commerce,
    Washington, D.C.

2.  Statistical Abstract of the United States, 1970, U.S. Dept. of Commerce,
    Washington, D.C.

3.  Guide for Compiling a Comprehensive Emission  Inventory, Environmental
    Protection Agency (June 1972).

4.  Systematic Study of Air Pollution  from Intermediate  Size  Fossil^ Fug!_
    Combustion Equipment, Wai den Research Corp.,  Cambridge, Mass.  (1970 •

5.  Communication with Mr.  Griffith,  Department of Statistics,  American  Gas
    Association, Arlington, Virginia.

6.  Climatological Data, Monthly Summarized Station and  Divisional  Data,
    U.S. Dept. of Commerce, National  Oceanographic and Atmospheric
    Administration, Ashville, North Carolina.

7.  1970 Census of Population - General Social  and Economic Characteristics,
    U.S. Dept. of Commerce, Washington, D.C.

8.  1970 Census of Housing - General  Characteristics, U.S.  Dept of Commerce,
    Washington, D.C.
                                    A-18
-------
                               APPENDIX B
                  REGRESSION ANALYSIS OF COMMERCIAL USE
                    OF OIL FOR VARIOUS SUBCATEGORIES
B-'l  GENERAL
     It was found that the commercial  category consisted of such a great
variety of business enterprises and institutions, that it would be in-
accurate to use employment as the distributive factor to allocate total
commercial fuel oil by county.  It was therefore decided to analyze the
relationship between fuel oil use and employment for several  subcategories
in order to determine the fuel oil use for these categories in a direct
way, based on the number of employees in each subcategory in  each county
(1).

     Fuel oil consumption of individual companies and institutions was
extracted from the NEDS files and basically compared to employment data.
This initial analysis has indicated fairly good correlations, and has
reconfirmed the fact that the various subcategories use fuel  oil in
very different ways.  The independent variable in these single variable
regressions was employment, but the slope and intercept of the regres-
sion lines differed considerably.

     It is recommended that this analysis be refined and pursued for
other categories as well.  The initial results are discussed below.

B-2  HOSPITALS

     The NEDS files provided Walden with a significant sample of hospitals
using fuel oil in Baltimore, New Hampshire and Massachusetts.  Staff data
on  hospitals were obtained from the American Hospital Association (2).

     For the 14 hospitals in Baltimore, fuel oil use was correlated with
an  R^ of  .91 to employment.  For the 15 hospitals in New Hampshire, the
  O
R  was  .86, and for the 40 Massachusetts hospitals (excluding state hos-
              f\
pitals), the R  was .68.  Data for all three states were combined, resulting
in  a correlation coefficient of  .81 and the regression line:

      fuel oil use  (thou. gallons) =  .715 x employees (thou.) + 208.5
                                   B-l
-------
B-3  SCHOOLS.

     Fuel  oil  use was available from NL'DS  for schools  in  Baltimore and
Massachusetts  and was compared to employment data  for  the specific schools
(3, 4).   The correlation coefficient, R2,  was .44  for  Baltimore and .58
for Massachusetts.  Due to the fact that many of the Baltimore schools
used fuel  oil  as a secondary fuel, it was  decided  to use  the  Massachusetts
results  based  on a sample size of 17.  The regression  line was:

     fuel  oil  use (thou. gallons) = 2.97 x employees (thou.)  + 76.2

B-4  COLLEGES

     Fuel  oil  use for colleges in Massachusetts, New Hampshire and Maine
was available from NEDS.  Staff data for these colleges were  obtained
from an HEW survey of employees in institutions of higher learning (5).
                                       2
     The Massachusetts data showed an R  of .76 for 15 sample points.
The New Hampshire and Maine data were combined and showed an  R2 of .65
for a total sample of 10 colleges.  Finally, the data  for all three states
were combined, producing an R2 of .67 and the regression  line:

     fuel oil  use (thou. gallons) = .546 x employees (thou.)  - 40.9

B-5  LAUNDRIE'S

     Fuel oil use for laundries available for Baltimore and Boston was
correlated with employment (6).  Originally, very poor correlations were
obtained.  Walden analyzed the Baltimore data, and found that many
laundry store fronts were included in the computer listing of  sources  in
Baltimore.  Walden proceeded  to separate out all the dry cleaning plants,
SIC 7211, 7216 and 7217, and  re-ran the regression analysis on dry cleaning
plants only.  An  R2  of  .72 was obtained for a total of 8 sample points.
It is obvious that more  data  need  to be collected for this subcategory.
In spite of this, it was decided  to use the resulting regression  line:

     fuel oil (thou. gallons)  =  .355 x  employment (thou.) + 8.4
                                  B-2
-------
     Walden was not able to obtain employment data for the various hotels
reported as point sources.   Therefore, a roundabout method was developed
to derive fuel oil consumed by hotels from employment data.

     Fuel oil use by hotels obtained from NEDS and from the public relations
offices of the Hilton Hotels was compared to the number of rooms for those
hotels (7), resulting in an R^ of .96 for a total of 18 sample points.
Then, the number of hotel rooms in each state was  correlated with the
number of hotel employees in each state (8) resulting in an Fe of .98.
State by state rooms/employees ratios were developed (see Table B-l) and
the resulting model consists of the regression line:

     fuel oil (thou. gallons) = 1.09 x rooms +41.5

where rooms are determined by dividing the employment in hotels in each
county by the appropriate state ratios.

Example:  To demonstrate the proportion of fuel oil used by these sub-
categories and residential apartment buildings, Walden calculated the
fuel oil used in Massachusetts by the enterprises and institutions dis-
cussed above, using the statewide coal, distillate oil, residual oil and
gas distribution shown in Section III of this report.  The results are
shown in Table B-2.
                                  B-3
-------
                TABLE B-l



ROOM/EMPLOYEE RATIOS FOR HOTELS BY STATE
State
AL
AK
AZ
AR
CA
CO
CN
DE
DC
FL
GA
HI
ID
IL
IN
10
KA
KY
LA
ME
MD
MA
Ratio
.28
.21
.29
.39
.27
.29
.37
.13
.58
.34
.33
.56
.21
.32
.30
.28
.23
.30
.38
.16
.29
.34
State
MI
MN
MS
MO
MT
NB
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
SD
TN
TX
Ratio State
.27 UT
.35 VT
.28 VI
.30 WA
.20 WV
.26 WI
.80 WY
.23
.27
.23
.31
.22
.21
.31
.22
.29
.36
.36
.28
.16
.28
.32
Ratio
.22
.26
.29
.26
.29
.31
.20















                  B-4
-------
                    TABLE  B-2

1970 COMMERCIAL USE OF FUEL  OIL  IN  MASSACHUSETTS
                 (Thou.  Barrels)

Hospitals
Schools
Colleges
Laundries
Hotels
Apartments
Z 6 categories
Total Commercial
Distillate
449
299
221
35
274
1>802
3,080
14,219
Residual
663
441
326
51
404
809
2,694
19,482
                       B-5
-------
                         REFERENCES -  APPENDIX  B

1.  County Business Patterns,  U.S.  Dept.  of Commerce,  Washington,  D.C.
2.  American Hospital Association Guide to the  Health  Care  Field,  Ameri-
    can Hospital Association,  Chicago, 111., 1970.
3.  Annual Report of Faculty Racial Composition,  Baltimore  City Public
    Schools, Office of Research Reports and Statistical  Records, Center
    for Planning, Research and Evaluation, Sept., 1972.

4.  Obtained from files of the Department of Education of the Common-
    wealth of Massachusetts.
5.  Number and Characteristics of Employees in  Institutions of Higher
    Learning, 1966-1967, Dept. of Health, Education and Welfare, Wash-
    ington, D.C.

6.  Dun & Bradstreet computer print-out provided  to Maiden  by the EPA.
7.  Hotel and Motel Red Book, American Hotel and  Motel Association,
    New York, N.Y.
8.  Hotels, Motor Hotels and Motels, Census of Selected Service Industries,
    1967 Census of Business, U.S. Dept. of Commerce, Washington, D.C.
                                   B-6
-------
                               APPENDIX  C
             REGRESSION ANALYSIS OF URBAN  VS  RURAL  DRIVING
                  PATTERNS ON A COUNTY-BY-COUNTY  BASIS
     Stepwise regression analyses were performed  on  county-by~county data
to find a relationship between miles traveled in  each county per registered
car and the socio-economic factors and highway patterns for each county.
The independent variables collected for this purpose were:   percentage of
population considered rural, miles of interstate  highways,  per capita in-
come, miles of state highways, and population density.

     The percent of county population considered  rural was  obtained from
the 1970 Census of Population (1).  The number of miles of  interstate high-
ways in given counties was obtained from the Federal Highway Administration
(2).  Per capita income by county was obtained from  the Department of
Commerce (3).  State highway mileage by county was obtained from the
California Highway Department for California only (4).  Population density
by county was obtained from the 1970 Census of Population (1).  Miles
traveled per registered car for each county was derived from total county
vehicle mile figures, provided by the highway departments of the states in-
volved and county car registration totals supplied by R. L. Polk of Detroit,
Michigan.  The results of the regression analyses are summarized in Table C-l
below.

     The results were very discouraging.  Not only were the multiple corre-
lation coefficients low, but the  intercepts of the regression lines and the
standard errors were extremely large.  In all states, there are some counties
with very extreme patterns.  When we eliminated the  data for the six counties
in California, which showed over  30,000 miles traveled a year per vehicle, a
                    o
greatly  increased R  (from .297  to  .440) and a sharply reduced standard
error  (from  9423 to 2377) were observed.  Similar changes are observed when
some of the  extreme data are eliminated in other states.
                                  C-l
-------












































"~J~
o

LU
_J
CO
•=c
r—



































































•z.
o
1 — 1
h-

V— 4
1—

UJ
o:
LU
[-1-
• — • 22:
o 2
LU
LU I—

t—i C£
S ex
OC
LU O
1— 'Z.
LU v- <
O >
O 5
h- Q

CO ^
t— 1 «c£
CO CO
>- C£
t — \
<£
z: co
^C ^"

•z. _J

»— * rv*
co rs
co o:
LU
o: z:
CD i—i
ill
DC:

u.
o

CO
i—

ra
CO
LU























1
S- +->
CU Q.
4-3 QJ
C 0
I— i
(O 4->
> rO
4->
"O to
C 1
CM •*->

• ~a
i. CU
rO 4->
>• O
QJ
-a i —
E QJ
CM CO

• TD
S- QJ
ro -t->
>• O
QJ
-a i—
C QJ
CM CO







»
^- •
ro 4-3
:=> ro
4J
•»-> in
to I
r— 4-^


•
S- .
ro 14-
>• M-
QJ
4-> O
 O
at
•4-* *~~




CO
CM
^— •


fx»
CTl
CM


^-^
00
in
«N 	 '

•
1)
t,—
f—
rO
O
10
«^J"
LO
in


UD
r*x.
en
•
CM





VO
«^-
^«

10
OJ
E

.
4J
to
QJ
^_
QJ
-l->
C






CO
r^
CO
•
^}"





<>o
CO
r—






r—
ro
S-
3
^_

S-5





o^
IO
r—
A
r*^



CTl
en



r-^
CO
CO



f, — ^
en
CO
^_^«

.
^:
in
rO
3:
CM
CM
i£>
f^


O
O

•
•3-





CO
oo


to
QJ
'g

.
4_j
to
QJ
S-
cu
+J
c:






o
CM

•
^^





CM
r—
r—






t— •
ro
S-
3
i-

6«





in
CD

n
«5^-



Ln
o
CO


en
r-.
CO

^-^.
o
o
r—
•* — *

to
rO
to
c~
ro
i^
O
cr>
t-.
I-,.


f^.
in
CO
•
UD





oo
f^N.
fM-

to
QJ
E

.
^_>
to
cu
£_
cu
4^
C






en

CO
.
UD





00
O







r—
rd
S-
3
L

&«





O
^J-
oo
•*
CO



00
«^
CO


m
r—
•3-




*~^*
en
en


ro


»— i
, —
CM
cr>
in



CM

,
m
oo




CO
en


in
cu
E

.
4J
to
CU
S-
QJ
-M
C





O

<^J-
un
•






en
CM
CM






f— •
rO
i.
3
j^

S^





CM
VO
r^*.
*i
en



^t

CM
	
O
o

* — "

ro
•i —
a>

o
cu
CD
en

•*
j
•~

en

o
.
f~~
.




t-^.
CM


in
cu
E

.
4_>
to
QJ
t.
QJ
4J
C






CTl
en
en
.
r-—





to
in







r—
ro
^_
3
i_

&-S





p>^
in
CM

CM



0
CM



r—
00
^



s~^
IO
r —
% — '

QJ

•r—
ro
2:
C3
CD

cn


r^^
) — .
CD
,
^j-





cn

i^

ifi
QJ
E

•
4_>
to
CU
t-
QJ
4_>
C






CM

r—
•
in





CO
CM
^~






r^
(0
^,
3


(?9





VO
co
in
M
^*



CM
«3
^-»


CD
O
CO
^^
m
i^.
^^^*

to
rd
1/1
C.
rO
^^
S-

«=*•
^j-
«=)•
, 	




1
1






1
1


-o
cu
4->
ra
• f—
E

f—
cu

a.
cu
•«-»






en

r^
•
CM
1

i —
oo

i



>>
4«>
•f—
to
r~
O)
-a

•
a.
o
Q.





cn
*^-
10





VO
J^I


\
^-
0

*
^— ^
f*^
(£>


ro
-a
• r—
s-
o
r—
U.























•
"O
QJ
S-
C1J
4J
in
•r—
cn
QJ
$-

^,
fC5
U

S—
OJ
O-

l/>
(U
r"^
•t —
s

4-
O

TD
rO
CU

(/I
C
•r—

-o

i_
cu
-(->
t/1
cn
QJ
S-

j_
rO
CJ

S-
QJ
a.
c:
O
^
cn

w
•r-
QJ

n
•2
rO
>«

4->
C
QJ

c~
QJ
Q.
QJ
Q
*
C-2
-------
     When a large percentage of the sample points  have to be eliminated
in a regression analysis, the significance of the  regression is sharply
reduced.   In Georgia, for example, the miles traveled per car range from
3167 for Chattahoochee County (27% rural)  to 56,976  miles for Charlton
County (100% rural).   Furthermore, for counties classified by the census
as 100% rural we find very dissimilar miles travelled per car (see Table
C-2 below).  This variability is probably due to factors relating to the
geographical location of the counties with respect to industrial regions;
factors which are not easily quantified.

                              TABLE C-2
     DRIVING PATTERNS FOR SELECTED 100% RURAL COUNTIES IN GEORGIA
County
Atkinson
Baker
Brant! ey
Calhoun
Charlton
Clay
Miles/Car
37,229
27,921
48,170
23,579
56,976
20,701
Interstate Mi les
0
0
0
0
0
0
Per Cap.
Income
1,828
1,696
1,820
1,875
1,988
1,745
     In short, the above analyses showed too vague a correlation between
driving patterns and rural/urban factors to be implemented in this pro-
ject.  It did show that the vehicle miles per registered car fluctuate
tremendously from county to county, giving us a rough idea of the extent
of our estimation errors in countywide gasoline consumption when car
registration figures are used for distributive purposes instead of the
preferred vehicle miles figures.  It is reassuring to note that many states
have now started collecting vehicle-miles data on a county level.  It would
                                   C-3
-------
be of interest to the EPA to indicate the usefulness  of these data to
the proper state highway department officials,  in  order to avoid  having
to use car registrations to distribute statewide gasoline use over the
various counties.

     To compensate for these errors to some extent, Walden summarized
the average miles driven per vehicle by four categories:  25% rural or
less, 26 to 50% rural, 51 to 75% rural and more than  75% rural, leaving
out the extreme values that varied by more than 50% from the mean of the
total sample in each state.  The results are shown below in Table C-3.
                              TABLE C-3
           AVERAGE MILES PER VEHICLE BY % RURAL CATEGORIES
State
California
Washington
Kansas
Iowa
Georgia
Maine
Arkansas
Seven State
Indexes
Entire
Sample
15,152
18,722
16,155
15,643
23,140
15,261
18,652
Avg.

Modified
Sample
13,047
17,092
15,179
14,604
20,262
15,261
17,932
16,709
100
_<25%
10,729
11,940
10,706
11,468
13,185
11,210
11,039
11,217
67
26-50%
12,202
15,793
13,622
12,141
17,680
15,396
15,908
14,572
87
51-75%
14,803
15,778
15,306
15,501
19,687
14,876
18,144
17,238
103
76-100%
16,920
21 ,855
17,225
15,862
22,461
15,995
19,458
19,261
115
     The  method  for  countywide allocation will be as follows:

     (1)   If  vehicle miles  are available by county, use vehicle miles for
 distributive  purposes  on  statewide gasoline for LDV figures.
                                   C-4
-------
     (2)  Otherwise multiply the LDV registrations in each county by the
appropriate miles/vehicle index according to the percentage of the popu-
lation considered rural in that county.

     (3)  Summarize these results for all counties to obtain a state total

     (4)  Use the modified county registrations, thus derived, to distri-
bute the statewide gasoline for LDV figures.
                                    C-5
-------
                    REFERENCES - APPENDIX  C
1970 Census of Population,  Characteristics  of  the  Population,  Number
of Inhabitants, U.S. Dept.  of Commerce,  Washington,  D.C.

Computer listing obtained from the Interstate  Reports  Branch of the
Federal Highway Administration, Washington, D.C.

Computer listing obtained from the Regional Economics  Division of the
Department of Commerce, Washington, D.C.

Historical State Highway, County Road and City Street  Statistics,
1957-1970, State of California, Department of  Public Works,  Division
of Highways, December 1971.
                               C-6
-------
                                APPENDIX D
      ANALYSIS OF AN ALTERNATE METHOD TO ALLOCATE  MOTOR FUEL USED BY
             HEAVY-DUTY VEHICLES ON A COUNTY-BY-COUNTY BASIS
      An alternate method to allocate gasoline and diesel  used by heavy-
duty vehicles to counties was analyzed.   It must be emphasized that this
method was not incorporated into the computer program and that data were
not collected for this method for 1972.

      The Federal Highway Administration provided state-by-state totals
of vehicle miles traveled by trucks on interstate urban and rural highway
systems.  It also provided estimates of the percentage of all trucks on
interstate highways which are diesel, distinguishing between urban and
rural interstate systems.  By applying these respective percentages to
the urban and rural interstate highway truck vehicle mile totals for each
state, we derived estimates for total vehicle miles driven by diesel trucks
on both urban and rural interstate highways in each state.

      Using county level interstate urban and rural mileage totals, also
provided  by the  F.H.W.A., the state vehicle-mile totals can be allocated
to counties.  Dividing the county diesel interstate vehicle mile totals by
5.1 MPG  (1), we  arrive at an estimate of total gallons of diesel fuel con-
sumed by  trucks  on interstate highways on a county basis.  This would
be  in the form of a rural and an urban subtotal.

      A  similar  procedure can be employed to estimate  gasoline consumed
by  heavy-duty vehicles on interstate  highways.  Using  this method, it is
first assumed that all heavy-duty vehicles not using diesel, use gasoline.
                                   D-l
-------
This is justified, since less than .1% of the truck traffic on interstates

is propane or LPG.  It is therefore assumed that total  interstate highway

truck vehicle miles minus the diesel  vehicle mile total  will be the gasoline

truck vehicle mile total.  This statewide gasoline vehicle mile figure is

then allocated to the various truck weight classes by using truck registra-

tions by weight class (2), adjusted to reflect the differing average miles

driven per year by trucks of different weight categories (3).  At this

point, the vehicle mile total for trucks weighing less than 6,000 Ib  (LDV)

is eliminated from further calculations.  The weight class vehicle mile

subtotals can now be divided by their respective mile per gallon estimates

which have been derived by Walden previously  (see page 3-15).  The resulting
  t
weight subtotals of gasoline consumption can now be combined and allocated

to counties  according to interstate mileage in each county, as in the

diesel method.


      The above methods  provide gasoline and diesel totals by county  con-

sumed by heavy-duty vehicles on interstate  highways.  It is  necessary to

allocate the remaining motor fuel, which is not consumed on  interstate

highways, to the  various counties  as well.


       1.  Diesel
           a.   Subtract the  interstate  diesel  consumed  in  the  U.S.  from
               the total  diesel  consumed on  highways  in the  nation.

           b.   Allocate the  resulting diesel  figure to  each  state,  by
               using the state percentage of the  original  national  total.

           c.   Allocate the  state diesel figure obtained in  Step b  to
               the various counties by  means of total truck  registrations.
                                   D-2
-------
The total diesel  consumed by heavy-duty vehicles in that county is the sum

of the diesel  consumed on interstate highways and the results of Step c

above.


      2.  Gasoline


          a.  Subtract the heavy-duty gasoline use on interstate highways
              from the total use of heavy-duty motor fuel in that state.

          b.  Subtract the state diesel consumption total (both inter-
              state and non-interstate) from the state total resulting
              from Step a.  This gives the consumption of gasoline by HDV
              on non-interstate roads in each state.

          c.  Allocate the state gasoline total obtained in Step b to
              the various counties by means of total truck registrations.


The total  gasoline consumed by heavy-duty vehicles in that county is the

sum of the gasoline consumed on interstate highways and the results of

Step  c above.  These modifications should eliminate the occurrence of

negative totals for county heavy-duty vehicle gasoline and diesel use,

which could occur in counties with high interstate mileage and small

truck registration totals.  The results of these new methods applied to

the thirteen test counties for the year 1971 are shown in Table D-l,

and can  be compared with  the results using the existing methods for the

same  year in  Table D-2.


      This method represents a significant improvement over the method

used  so  far by Walden to  allocate gasoline and diesel use by county.   It

is found to have the following drawbacks, however:
                                    D-3
-------
a.  It is a very time-consuming method with regard to the



    required input preparation.



b.  The required data are not presently available on a yearly



    basis.
                              D-4
-------
































1
Q

LjJ
, 	 |
CO
«t
(—















































U:
Q
O
:r
r—
LU
s:

LU
	 1
t — i
5:
i ' t
o
fc-H
LU
>
LU

1—
to
LU
) — » — •*
"Z. CO
O
CD r—
K-< ro
to CD

f—
i — O
cr> to
i— "O
C.
Qi ro
O to
1 1 ^
O
1— t—

ZD
to
LU
C£
2^
O
i — i
1—
to
p**
8

	 i
LU
U_

f—

;— ^
o
o











>>
i — n c
4-> 1 i —
o -i-1 o
I— -c co
en ro
•r- CD
_J
ro O>
4-> CO
o cu
5



OJ
r— fd
O) 1 4->
CO C CO
QJ O *-
•r- Z 0>
a 4->
C
t — 1

4-»
i — (O
aj +->
CO CO
ni SM

Q -S
C
1 — 1



•4-> a>
3 C
r- Q-r-
rO 1 r—
I \ ^^^ .p^
•*-* ^) O
O > co

OJ CD
"T~



>)  a; -t->
^ c fd
Q •<- 4-»
1 i — CO
*~ .-. *^
>> o >-
> co O)
rO ra 4->
OJ CD C
3= •-

1
if
Jv O fl^
_l ji ^/f i -.*
9 (O
Q 41 4J
1 C! (f>
^
4J
C
3
O
O


CT» tO i — «3" OJ
^f 00 ^ ««- LO
OJ OJ 00 ID OJ
00 r^. CTt to tO
OJ O LO r—
•»
oo
LO gj ^ g j-
o> o oj «3- sf
r— OJ CT\ f — *£>
r— 00 OJ




vo «* r-~ <^- to
LO r^ oo c> CD
V£) OJ OJ «3" r—
LO ^— *3* i — • OO
tO r—


O"l tj t^1^ t!I^> U^
^^ ^^ O^ C^ to
OJ OO tT> CO
vo o «*• co
OJ r—





r^ o co o r—

LO OO CO CO r~~
tn to to LO LO
OJ OO LO i —
r—






OO r"" LO CJ f^
CO ' r"^ LO r^ CO
f~~> tO CO r"~*
OO LO i — i —
00 OJ





«3- CT> 00 O r—
r*^ oJ oj LO oj
CO t^. O CO LO
tO O LO LO ^*
OJ LO OO i —








to
0} f"*
C r- 0 0

co cr> fu i- tl)
 LO i — LOOl —
I^-LO^OI^-OOLOO
oooj«^-r— tor— cnr--
O CSJ OJ <^~ r— LO to OJ
LO OJ i — •*


C\J *^" I**1* CO €^f ^*^ C\J VO
o^ csj ro LO o^ f~^ i^^ r^**
O^ LT> vO O^ *«1" f^"* CO
r— CO OJ




CD ^J" VO CD ^" r— • LO Ovf
co r*^ ^j* ^o LO o^ r^* LO
o>v0ojcomcr»coo
^£> ^- i — F— cr»



LOCXJ^O^LOOOU}
^t' r*^ oj oj r^» co LO P—
cr> LO ^^ r*^ LO o co r*^
oj ^3* oj r^. co LO en
i— OO •—





•— OOOljOLOr— 0>
i— i — ojr~-otOLocT)
OJOJi— OO'eJ-COLOr—
OO ^J- QO OO r*~ *^* LO LO
i — 1 — ( — ^3*







oo f^^ t7^ oj co oo r^^ to
COtOi — r — OCOLOLO
r— OlOCT>^CJ>«3-tO
i — OJ «3" OJ ' — LO
r—





COOOr— COCOOJ«J-OO
r^> <^* co LO CTI r^ cr> ^j*
Oi — LOO-ltDOOOLO
r^ oj oo LO co «s- cri
i— r— OO




\
>> c
-M 3
•r- O
0 CJ

-------
                                TABLL U-2

   COUNTY CONSUMPTION OF MOTOR FUELS BY HEAVY AND LIGHT-DUTY VEHICLES
                     USING EXISTING METHODS* - 1971
County
Jefferson
Los Angeles
San Diego
Boulder
Palm Beach
Baltimore
Worcester
Franklin
St. Louis
Bel knap
Minnehaha
Galveston
King
Heavy-Duty Vehi
Gasoline Consumption
29,960
154,800
34,860
6,035
17,250
34,190
18,580
2,750
10,400
1 ,030
2,466
5,722
32,880
cle
Diesel
Consumption
15,790
171,300
39,800
4,240
8,674
19,490
310,800
26,470
5,194
987
2,765
503
24,040
Light-Duty Vehicle
Gasoline Consumption
181,400
2,731,000
499,000
55,120
143,900
467,900
11,860
1,883
87,070
19,500
25,230
71 ,080
469,100
*See page
                                   D-6
-------
                                   rrCHNICAL HI;KM I DATA
                               ' rcc.d iHiiruamri'S on r/i.p revrrr:C /"•/"'<' (
I  (' i t'O.I i NO

. _£ PA^i507_3z7 4^0.2.1	
-;  1 I TLE AND SULJTITLfc

 Development  of a  Methodology  to  Allocate Liquid
 Fossil  Fuel  Consumption by County
                                     5. PERFORMING ORGANIZATION CODE
                                                            HhCII-'ll Nt f. ACOuSSIOONO.
                                            DAI F
                                      March  1974
7  AUTHORIS)
                                                           a. PERFORMING'ORGANIZATION REPORT NO.
 Josette  C.  Goldish, Franklin  D.  Trowt, John R.
 Fhrpnfpld,  Khee M. Chna.  Richard Stockdale
-"> PcRFORMING~OR'~.ANIZATION NAME AND ADDRESS

  Wai den  Research Division of Abcor Inc.
  359  Allston St.
  Cambridge, Massachusetts   02139
                                     1O. PROGRAM ELEMENT NO.
                                     11. CONTRACT/GRANT NO.
                                        68-02-1067
 2. SPONSORING AGENCY NAME AND ADDRESS
 U.S.  Environmental Protection Agency
 Office of Air Quality  Planning and Standards
 Research Triangle Park,  N.  C. '27711
                                     13. TYPE OF REPORT AND PERIOD COVERED

                                                               1974 .
                                     14. SPONSORING AGENCY CODE
 5. SUPPLEMENTARY NOTES
 6. ABSTRACT
       Methods were  developed for the routine determination of distillate  and
  residual oil consumption by industrial, commercial,  and residential consumers,
  as well as for gasoline and diesel fuel consumed  by  light and heavy duty motor
  vehicles.  The resulting data are allocated to  counties for input and  storage
  in the National  Emissions Data System  (NEDS)  area source format.  In addition,
  seasonal fluctuations of fuel oil use  by consumer category and geographic region,
  and references for determining sulfur  content of  fuel oils on a county basis,
  were analyzed.   The report summarizes  the methodologies that were developed
  and describes the  computer processing  techniques  for reporting the data.
                                KEY WORDS AND DOCUMENT ANALYSIS
                   DESCRIPTORS
                        b.lDENTIFIERS/OPEN ENDED TERMS ' |c. COSATI Field/Croup
  Fuel consumption
  Fuel oil
  Gasoline
  Diesel  fuel
  NEDS
  Area source
Seasonal Fluctuations
Sulfur content
Counties
  g. DISTRIBUTION STATEMENT

   Release Unlimited
                         19. SECURITY CLASS (ThisReport)
                         	Unci_a_s_s i f i e d
21. NO^ OF PAGES

   140
                                               20 SECURITY CLASS (This pa?
                                                 Unclassified
                                                   22. PRICE
 EPA Form 2220-1 (3-73)
-------
-------