United States
                    Environmental Protection
                    Agency
National Kisk Management
Research Laboratory
Cincinnati, OH 45268
                    Research and Development
EPA/600/SR-95/161
February 1996
& EPA      Project Summary
                    Site-Specific Characterization
                    of Soil  Radon   Potentials
                    Kirk K. Nielson, Rodger B. Holt, and Vern C. Rogers
                      The Florida Department of Commu-
                    nity Affairs is developing construction
                    standards for incorporating radon-re-
                    sistant building features in areas of el-
                    evated soil radon potential. Although
                    statewide maps have been developed
                    to show the regions where the features
                    are  required, there is also a need for
                    simple methods to assess the radon
                    potential of specific building sites. The
                    report gives results of the development
                    and evaluation of a mathematical basis
                    for using simple site measurements to
                    estimate soil radon potential. The ap-
                    proach utilizes  a lumped-parameter
                    model of radon generation and entry.
                    Site-specific soil radon potential is de-
                    fined as the rate of radon entry into a
                    reference house, consistent with previ-
                    ous definitions used for the statewide
                    radon maps. The model shows that, in
                    the simplest case, soil radon potential
                    is reduced to a simple function of two
                    measurable  parameters:  the soil sur-
                    face radon flux and the soil moisture
                    (as  a fraction of saturation). The flux
                    gives the radon generation rate of the
                    soil profile, and the moisture is a sur-
                    rogate for radon transport parameters,
                    including  air permeability  and radon
                    diffusion coefficient.
                      Field tests of soil radon flux and
                    moisture measurements were  con-
                    ducted at 26 house sites in Polk County,
                    Florida, to evaluate their utility in pre-
                    dicting site-specific radon potentials.
                    Radon fluxes also were measured from
                    bare concrete  surfaces,  where they
                    were accessible, to better estimate non-
                    advective radon entry rates. Gamma-
 ray intensity also was measured in the
 yards, but it failed to correlate well with
 the radon fluxes.  The measured soil
 radon fluxes and moistures showed lo-
 calized trends that compared well with
 mapped  radon potentials  in some
 cases, but  not in others. For the 26
 sites, the radon  potentials estimated
 from site-specific measurements aver-
 aged twice the potentials from the gen-
 eralized radon maps. A large  geomet-
 ric standard  deviation (GSD =  4.7) was
 associated with individual sites.
   The site-specific estimates also were
 compared with prior indoor radon mea-
 surements. When the reference-house
 ventilation rate was attributed to the
 houses, the calculated/measured radon
 ratios averaged 1.06  ± 0.72. Slightly
 greater bias  but improved precision
 (0.87 ± 0.56) was obtained using  con-
 crete-surface radon flux measurements
 in addition to the site  radon potential
 measurements. The empirical measure-
 ments  suggest that the  precision of
 site-specific  evaluations  is marginal,
 leaving an uncertainty  of about a fac-
 tor of 2 in site-specific estimates. Al-
 though potentially useful for some ap-
 plications, the site-specific measure-
 ments studied here do not greatly im-
 prove the radon potential estimates
 over the regional estimates already
 available  from radon maps.
   This project summary was developed
 by the National Risk Management Re-
 search Laboratory's Air Pollution Pre-
 vention and Control Division, Research
 Triangle  Park, NC, to announce key
 findings of the research project that is

-------
fully documented in a separate report
of the same  title (see Project Report
ordering information at back).

Introduction
  Radon (222Rn) gas generated by natu-
rally occurring  radium (226Ra) in soils  can
enter  buildings through their foundations.
With elevated  entry rates and inadequate
ventilation, radon can accumulate indoors
to levels that pose significant risks of lung
cancer with chronic exposure. The Florida
Department of Community Affairs (DCA)
and the EPA have jointly developed radon-
resistant  building standards to help re-
duce  radon  entry  from soils.  The stan-
dards  address improved  understructure
sealing, altered air pressures,  and other
engineered features developed under the
DCA's Florida  Radon  Research Program
(FRRP). Statewide radon potential maps
also have been developed to identify re-
gions where the radon-resistant features
are needed. This report examines the fea-
sibility of estimating soil radon potentials
from simple measurements to help deter-
mine  more accurately  the radon  protec-
tion needed at a specific building site.
  Soil radon potential is defined using the
same reference house as was used previ-
ously  for the maps. A  lumped-parameter
model simplifies the theoretical  basis,  and
indicates minimum parameters  and surro-
gates for characterizing the site radon po-
tential. Field tests of the methods include
measurements of selected parameters and
surrogates  at  26 house  sites  that were
already being  studied  under the  FRRP.
Soil radon  potentials estimated from the
field measurements were compared with
mapped regional estimates, and were also
compared with data from measured radon
levels in the houses at the sites.

Previous Estimates of Radon
Potential
  Radon indices and simple models have
been  proposed previously for  estimating
soil radon potential. These have depended
variously on house ventilation rates, ema-
nating  soil radium concentrations,  soil air
permeability, soil radon concentrations, soil
porosity, soil radon migration distance, soil
water permeability, and soil equivalent  ura-
nium concentration. A  more detailed ap-
proach numerically analyzes advective ra-
don transport into houses. This approach
utilizes soil air permeability, soil radon gen-
eration rate (radium  concentration,  den-
sity, and  radon emanation coefficient),
foundation crack geometry, and house air
pressures.  It  defines  radon  potential as
the rate of radon entry into a  house in
picocuries per second. A review  of  site
measurement  methods shows the need
for  detailed  radon source and  transport
measurements, including soil density, par-
ticle size, texture  classification,  moisture,
permeability,  diffusion  coefficient,  radon
emanation coefficient, radium concentra-
tion, and  radon  concentration  profiles.
These  properties are  used in  a  radon
source potential index that depends also
on  site drainage conditions,  site ground-
water conditions, and site climatology.
  A more  detailed  modeling  approach
characterizes radon entry from house and
soil  parameters,  including  radon move-
ment by both advection and diffusion.  Us-
ing  the RAETRAD model, this  approach
uses detailed soil  radium  distributions; ra-
don emanation fractions; and soil density,
moisture, permeability, and diffusion coef-
ficients with house air pressure and crack
distributions.  Soil  radon  potential is  de-
fined on an annual basis (in millicuries per
year) to emphasize the  long-term average
nature  of  equivalent steady-state  radon
entry rates and exposures.

Reference House
  The site-specific soil  radon potential is
defined as the annual rate of radon entry
from soils into a hypothetical reference
house that is defined to represent Florida
slab-on-grade  houses.  The  reference
house provides a constant, typical inter-
face between the indoor exposure volume
and the varied soil conditions that control
radon potential. Although house and  soil
parameters cannot be  completely sepa-
rated for modeling radon entry, the use of
a reference house avoids the large differ-
ences in radon potential that would other-
wise result from differences in house de-
sign, construction, ventilation, and occu-
pancy.
  The  present reference house corre-
sponds to the  house defined  previously
for  radon potential mapping. The house is
an  8.6 x 16.5 (28 x 54 ft), slab-on-grade
single-family dwelling. Its volume is based
on that of a  median U.S.  family dwelling,
and is similar to  that  of typical  Florida
houses. Its area is estimated from its vol-
ume using a nominal 2.4-m (8-ft) ceiling
height.  Its ventilation rate  is about half the
normal median U.S. house ventilation rate,
based on measurements in Florida houses.
A perimeter  floor crack  approximates a
floating-slab  shrinkage crack to permit ad-
vective radon entry from  pressure-driven
air  flow. The stem wall and  footing pen-
etrate 61 cm  (2 ft) into the natural terrain,
and enclose an additional 30 cm (1 ft) of
above-grade fill soil beneath the  slab. The
indoor pressure is -2.4 Pa, typical of pres-
sures from thermal and wind-induced pres-
sures in U.S. houses, and also  typical of
the  average  pressures measured in 70
Florida houses  under passive conditions.
Concrete slab air permeabilities, radon dif-
fusion  coefficients,  and other properties
are estimated  from data measured  on
Florida floor slabs.
  Soils beneath the reference house are
modeled  as uniform, isotropic soils with a
bulk density of  1.6 g cm"3. A 30-cm  layer
of sandy fill soil is located  beneath the
slab, below which the site-specific soil  is
represented by its textural class  and  its
associated  water content at a matric po-
tential of -30 kPa. From these properties,
the soil air permeability and radon  diffu-
sion coefficient are calculated from em-
pirical relationships.

Lumped-Parameter  Model
  The  mathematical definition of site-spe-
cific radon  potential utilizes a lumped-pa-
rameter model,  which is based in turn  on
the detailed  RAdon Emanation  and
TRAnsport  into Dwellings  (RAETRAD)
model. The lumped-parameter model was
developed  primarily from  RAETRAD sen-
sitivity analyses, which identified the  most
significant house and soil parameters. The
analyses suggested simplified approxima-
tions to express average indoor radon lev-
els as a function of radon source strength
and house  radon  resistance and ventila-
tion parameters. Radon  source strength
was defined in  terms of the sub-slab  ra-
don concentration. House radon resistance
was defined from floor openings, pressure
driving forces,  and slab diffusivity.  The
lumped-parameter  model  uses  a  simpli-
fied relation between indoor radon and
the radon entry rate:

Cnet=Cin-Cout =3.6Q/(XhVh)  (1)

where   Cnet  = net indoor radon concen-
               tration from sub-slab
               sources (pCi L~1)
        Cin  = total  indoor radon  con-
               centration (pCi L1)
        Cout  = outdoor background radon
               concentration (pCi L1)
        3.6  = unit conversion (pCi L1 rr1
               per pCi rrr3s"1)
        Q    = radon entry rate (pCi s"1)
        Xh    = rate of house ventilation
               by outdoor air (rr1)
        Vh   = n Ah = interior house vol-
               ume (m3)
        h    = mean height of the  inte-
               rior volume of the house
               (m)
        Ah   = house area (m2).

  The  radon entry rate in equation (1)  is
defined in  the  lumped-parameter model
for the reference house as:

-------
     Q = AhCsub[fc(vdc-vacAP)

        + Vs|ab+VSc]               (2)

where   Csub  =   sub-slab radon concen-
                 tration (pCi L1)
        f,    =   area of floor openings
                 as a fraction of total
                 floor area (dimension-
                 less)
        vdc   =   equivalent velocity of
                 radon diffusion through
                 floor openings, depen-
                 dent on the radon dif-
                 fusion coefficient of the
                 soil (0.0143 mm s'1)
        vac   =   equivalent velocity of
                 radon advection through
                 floor openings, depen-
                 dent  on the air per-
                 meability  of the  soil
                 (mm s'1 Pa'1) = exp(-
                 3-0.045e6S)
        S    =   soil water saturation
                 fraction (dimensionless)
        AP   =   indoor air pressure (Pa)
        vsiab  =   equivalent velocity of
                 radon diffusion through
                 the slab, dependent on
                 the  radon   diffusion
                 coefficient of the slab
                 (mm  s-1)  = 2.9x10-7
                 exp(11.4W)
        W   =   water/cement ratio of
                 the slab concrete (dimen-
                 sionless)
        vsc   =   radon entry velocity ad-
                 justment for house size
                 and crack  location (mm
                 s-1) = 3.5x10-=(xcrk/xh)
                 + 4.6x10-5/xh
        xcrk   =   location of dominant
                 floor crack opening from
                 house perimeter (m)
        xh   =   house width (m).
  Only  the  parameters Csub and  vac in
equation (2) are site-dependent; therefore
reference-house values  were substituted
for all of the others, leading to a simplified
relationship for defining the  site-specific
soil radon  potential:
Qss =1.68Csub [0.019

      + exp(-3-0.045 i
                                   (3)
  where Qss = site-specific soil radon po-
tential (pCiV1).
  Although the indoor radon concentra-
tion for a reference house can be directly
estimated from Qss in equation (1), indoor
radon concentrations for specific houses
are better  estimated by using  as  many
defining parameters in the lumped-param-
eter model  as are known. Using the defi-
nitions associated with equation (2) to de-
fine the radon entry rate for equation (1),
indoor radon concentrations  can poten-
tially be better estimated from house-spe-
cific data.

Surrogate Estimates of Model
Parameters
  Many model parameters  are difficult to
measure directly, and therefore are  sel-
dom quantified.  However,  most can  be
estimated  from  related  parameters that
are directly measurable. The site-specific
value for the soil water saturation fraction
(S) can be  readily estimated from mea-
sured soil  moisture contents as:

       S = 0.01 Mv/e = 0.01 pMw/e    (4)
where   M  =
        P   =
        M  =
 soil moisture (volume per-
 cent)
 total soil porosity (dimen-
 sionless) = 1 - p/pg
 soil bulk dry density (g
 cnr3)
 soil specific gravity (nomi-
 nally 2.7 g cnr3)
 soil moisture (dry weight
 percent).
  Using the reference-house slab param-
eters and assuming that the radon-gener-
ating soil  profile is deep (unconstrained
by a water table or bedrock),  Csub can be
approximated as:
   Csub = (90 + 5,900JS)

          /(1.13 + 35 ,/DJ
where   J  =
        D  =
                     (5)
radon flux at the soil sur-
face (pCi rrr2 s'1)
radon diffusion coefficient
of the soil pore space (cm2
s-1).
Field Tests
  Sensitivity  analyses  with the  lumped-
parameter  model have demonstrated  a
relatively strong dependence of radon po-
tential on Csub,Xh, W, AP, S, h, and f., and
a smaller dependence on vdc, xh, xcrk. Field
measurements therefore were directed at
quantifying  the important parameters.
Where the parameters could  not be ad-
equately measured,  default values typical
of the reference house were used.
  Site-specific field  measurements  were
conducted during March  17-22,  1993, to
evaluate the sensitivity,  precision, and util-
ity of selected parameters  for estimating
site-specific radon potential. The  measure-
ments were conducted  in the yards  of 26
houses  in Polk County,  FL, for  which in-
door radon  data were  already available.
                                                                                The protocol at  each site included  mea-
                                                                                surements on all four sides  of the house
                                                                                of soil moisture, gamma-ray activity, and
                                                                                radon  flux.  In  addition,  radon flux  was
                                                                                measured from  a  bare  concrete  surface
                                                                                where suitable locations were accessible.
                                                                                  The site protocol concentrated on rapid,
                                                                                inexpensive  measurements that could
                                                                                most directly estimate the site-specific ra-
                                                                                don potential, Qss.  As indicated by equa-
                                                                                tion (3), the sub-slab radon concentration
                                                                                (Csub) and the soil water saturation fraction
                                                                                (S) were of primary interest. The soil mois-
                                                                                ture, Mv, was measured using a time-do-
                                                                                main reflectometer probe, which  charac-
                                                                                terized the top 30 cm of soil. The value of
                                                                                S was calculated from  equation  (4), as-
                                                                                suming a soil porosity of e =0.407.
                                                                                  Although prior FRRP  measurements of
                                                                                Csub were planned, these data were gen-
                                                                                erally unavailable,  and consisted of  only
                                                                                single measurements in a  few cases.
                                                                                Therefore, the soil surface radon flux mea-
                                                                                surements at each site became  the pri-
                                                                                mary estimator of Csub, using equation (5)
                                                                                as the basis. The flux measurements were
                                                                                made  using  the small-canister  method,
                                                                                which  gives  equivalent  results  to  EPA
                                                                                Method  115. The  radon  fluxes were
                                                                                sampled over a 24-hour period, after which
                                                                                the  charcoal  canisters  were retrieved,
                                                                                sealed, and submitted for laboratory as-
                                                                                say of radon activity. The  value of Ds
                                                                                required for calculating Csub was estimated
                                                                                from the same porosity and  moisture us-
                                                                                ing the predictive correlation:
                               Ds = D0 e exp(-6 e S - 6S
                                                      14e,
                                                                                                                   (6)
                                                                                where  Do =  diffusion coefficient for ra-
                                                                                              don in air (0.11 cm2 s'1).
                                                                                  The radon  flux measurements on bare
                                                                                concrete slab surfaces, when accessible,
                                                                                estimated more  directly the radon  entry
                                                                                through the intact portions of the concrete
                                                                                slabs. The flux measurements directly es-
                                                                                timate v  for use in equation (2) as:
                                                                                    v slab-Jslab/Csub                (7)

                                                                                               radon
                                                                                               crete slab surface (pCi nr2
                            where  Jslab =  radon flux from the con-
                              The small charcoal canisters were found
                            to have marginal sensitivity for the lower
                            fluxes from concrete surfaces. Therefore,
                            an alternative method used approximately
                            230  cm3  of  granular  activated carbon
                            spread over a paper napkin mounted in a
                            30-cm diameter wooden  compression
                            frame and covered with  a polyethylene
                            sheet. The frame was sealed  to the con-
                            crete surface  with rope caulk.  After a 24-
                            hour deployment, the  charcoal was  re-

-------
trieved and sealed  into metal cans  for
assay  of  radon.  The  method was cali-
brated  against the small-canister samplers
using a thin-sample radon  source. The
large-area method measured fluxes equal
to those from small canisters, but with 13
times greater sensitivity.
  The  gamma  ray measurements  were
intended  for possible correlation with the
soil radon flux measurements, or as po-
tential  surrogates for surface soil radium
concentration. The measurements  were
made 1 m above the soil surface  using a
5 x 5-cm sodium iodide scintillation probe.

Test Results  and Analysis
  The  site-specific measurements  were
first analyzed for simple empirical correla-
tions with the measured soil radon fluxes.
The measurements also were used to pre-
dict site-specific radon potentials, which in
turn were compared with the FRRP indoor
radon concentration data. The gamma ray
measurements showed  greater differences
among the different house sites than the
radon flux  measurements, which in turn
showed greater differences than the mois-
ture  measurements. All of the variations
among house sites were significant at the
p<0.01 level in analyses of variance.
  Since radon flux expectedly varies with
radium in  uniform soils, a linear flux vs.
gamma relationship was sought by least-
squares linear regressions.  The  regres-
sion  on individual  gamma intensity  mea-
surements (y in uR h'1)  gave a correlation
coefficient of only r =  0.26  for the fitted
line Js = -0.2 + 0.146 y. A regression corre-
sponding to Js = 0.23  y062 was  obtained
from log-transformed data. The regressions
were strongly affected  by some low flux
points at high gamma intensity that were
associated with wet soils. The high uncer-
tainty in the flux vs. gamma relationship
limits the  usefulness of gamma  intensity
as a surrogate for soil radon flux.
  A similar  regression  of soil radon flux
on soil moisture gave similar scatter. This
regression on individual  moisture measure-
ments had a correlation coefficient of r =
0.26 for the fitted line Js = 6.9 - 0.268Mv.
Log-transformed data gave the line Js =
28 M;118. Low fluxes were associated with
high soil moisture levels.
  Measured ra_don fluxes also were  re-
gressed on yV Ds-tanh(xsV  A/DS), where
X is the radon decay constant (2.1x10'6
s'1) and xs is the soil thickness dominat-
ing the flux (i.e.,  above the  water table).
Since y is a surrogate for radium concen-
tration, this lumped parameter is the theo-
retical surrogate  for radon  flux,  Jsur,  for
uniform  soil. The measured  radon fluxes
were regressed  on Jsur, which  utilized ra-
don  diffusion  coefficients from equation
(6)  and measured moistures. The regres-
sion had an improved  correlation coeffi-
cient of r  =  0.55 for the  fitted line Js  =  -
0.85 + 1.51 Jsur. Log-transformed data gave
the least-squares fitted line  Js = 0.6 Jsur.
Although this  relation better  predicts  ra-
don  flux, it still exhibits considerable  un-
certainty, which limits its potential  use in
predicting radon flux.

Estimation of Site Radon
Potentials
  Site-specific soil radon potentials were
estimated using  equation (3). Measured
soil moistures were used to determine S
from equation (4), and  measured  radon
fluxes were used to determine Csub from
equation (5). The calculated  site-specific
soil radon  potentials  were averaged by
neighborhood groupings for  area-based
comparisons with the mapped soil radon
potentials.  Figure 1 illustrates the  result-
ing geometric means and geometric stan-
dard deviations, and gives side-by-side
comparisons with the  median  mapped ra-
don potentials. For illustration purposes,
average  error bars also  were applied  to
the three houses not associated with other
neighborhood groupings. As illustrated, the
Qss values were higher than the mapped
radon potentials (Q   )  in six of the ten
comparisons. In statistical analyses of the
seven  comparisons  involving  multiple
houses, the Qss values averaged  0.8 stan-
dard deviations higher than the  Qm?  val-
ues.  This average bias  is not significant
(p<0.41). The geometric mean  of  all 27
ratios of Qss/Qmap was 2.02, with a geo-
metric standard deviation of 4.7  (GSD  of
the mean is 1.35). The positive bias in Qss
may result in part from radon flux sam-
pling near the houses, which can increase
                        102
                    1  10°
                    cc
                    "5
                    tn
                        io-1
                                                                                  Ir
                                                           D
                                                           Neighborhood

Figure 1.  Comparison of site-specific and mapped soil radon potential distributions for seven neighborhood groupings (A-G) and three isolated houses.

-------
the flux compared to an open-field sample
that is away from the house foundation.

Comparison of Soil Radon
Potentials with Indoor Radon
Data
  Indoor radon  concentrations were esti-
mated from the soil  radon  potential mea-
surements for comparison with the FRRP
indoor radon data. The indoor radon con-
centrations were estimated from equation
(1) using the  reference-house ventilation
rate  (Xh = 0.25  h"1)  with individual  esti-
mates of  house volumes.  The  ratios  of
calculated/measured indoor radon concen-
trations  averaged 0.94 ± 0.65 for  the 13
slab-in-stem-wall houses and 1.18  ± 0.80
for the  13 monolithic-slab  houses.  The
overall  average  ratio for all 26 houses
was  1.06  ± 0.72.  Despite  the  relatively
large scatter,  this  comparison demon-
strates close average agreement of calcu-
lated and measured radon concentrations.
  A separate comparison also utilized the
measured radon fluxes from concrete sur-
faces, Jslab. Using equation (7), the fluxes
defined  vs|ab, which was used in  equation
(2) with reference-house values for the
other parameters. The ratios of calculated/
measured radon concentrations averaged
0.97 ± 0.64 for the 10 slab-in-stem-wall
houses, and 0.69 ± 0.35 for the five mono-
lithic-slab houses where concrete surface
fluxes were measured. The overall aver-
age  ratio for  all 15 houses  was 0.87 ±
0.56,  compared to  1.18  ±  0.89 for  the
corresponding houses  by the  previous ap-
proach.  The  lower  variation  from  using
the concrete  fluxes shows that they  im-
prove the  estimate of  radon transport
through slabs over the generic assump-
tion  of equation (2). This approach has
slightly larger biases, but  significantly im-
proves the precision  over the previous
approach.

Discussion and Conclusions
  The analyses in this  report demonstrate
a theoretical basis for measuring site-spe-
cific  radon potentials  from a  variety of
potential  surrogate parameters. Although
radon flux and soil moisture were the  pri-
mary parameters evaluated,  others (e.g.,
soil radium, soil radon, soil air permeabil-
ity,  and  soil  density)  may  also  provide
useful results. The flux and moisture mea-
surements were chosen for their simplicity
and low cost.  However, their marginal pre-
cision leaves an  uncertainty of about a
factor of 2. This uncertainty can compro-
mise  the  basic purpose of site-specific
measurements, which  is to reliably deter-
mine the potential for elevated radon at a
particular building  site.
  The need for better precision and accu-
racy may require other measurements. For
example, soil radium  profiles measured
from borehole samples support more de-
tailed  model analyses that determine ra-
don potentials with greater precision and
accuracy.  Although  more costly, radium
profile measurements have demonstrated
model and measurement agreement within
10-20%.
  If site-specific measurements  are  con-
sidered for construction decisions, the plan-
ning should consider measurement costs,
radon control costs,  measurement uncer-
tainty, and  radon  control uncertainty.
Ample design safety margins should allow
for  measurement  and control uncertain-
ties. The safety margins also give benefits
of reduced health risk from lower radon
levels, thereby serving a greater purpose
than simply assuring attainment of a  < 4
pCi L1 indoor radon  goal.

-------
   K. Me/son, R. Holt, and V. Rogers are with Rogers and Associates Engineering
     Corp., Salt Lake City, UT 84110-0330
   David C. Sanchez is the EPA Project Officer (see below).
   The complete report, entitled "Site-Specific Characterization of Soil Radon Poten-
     tials, "(Order No. PB96-140553; Cost $17.50, subject to change) will be available
     only from:
           National Technical Information Service
           5285 Port Royal Road
           Springfield, VA 22161
           Telephone: 703-487-4650
   The EPA Project Officer can be contacted at:
           Air Pollution Prevention and Control Division
           National Risk Management Research Laboratory
           U.S. Environmental Protection Agency
           Research Triangle Park, NC 27711
United States
Environmental Protection Agency
National Risk Management Research Laboratory (G-72)
Cincinnati, OH 45268

Official Business
Penalty for Private Use $300
      BULK RATE
POSTAGE & FEES PAID
         EPA
   PERMIT  No. G-35
EPA/600/SR-95/161

-------