United States
              Environmental Protection
              Agency
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
EPA-454/R-93-055
October 1993
              AIR
& EPA
    AN EVALUATION OF A SOLAR RADIATION/DELTA-T METHOD
FOR ESTIMATING PASQUILL-GIFFORD (P-G) STABILITY CATEGORIES

-------
                                                                EPA-454/R-93-055
O
              An Evaluation of a  Solar Radiation/Delta-T Method
         for Estimating Pasquill-Gifford  (P-G)  Stability  Categories
                                   October 1993  U.S. Environncr.':! ' . -Action Asency
                                                  Region 5, Library •;;-•.-j>!)
                                                  77 West Jackson Boulevard, 12th Floor
                                                  Chicago, IL  60604-3590
                        U. S. ENVIRONMENTAL PROTECTION AGENCY
                     Office of Air Quality Planning and Standards
                             Technical Support Division
                          Research Triangle Park, NC  27711

-------
                   ACKNOWLEDGEMENTS

Special credit and thanks are due Dr. Desmond Bailey and
Mr. John Irwin for their technical assistance and advice
through all phases of the project,  from meteorological
data  acquisition  and processing through  testing  and
development of the SRDT methodology. Thanks are offered
Mr.  Gerry  Moss  and Mr.  Pete  Eckhoff   for  FORTRAN
programming support and in meteorological data retrieval
and reformatting.  Thanks are  offered to  Mr.  Russ Lee
and Mr. Roger Erode,  Pacific Environmental  Services for
their consultation on ISC2 runs.  Special appreciation
is due Mr. Jim Paumier,  Pacific Environmental Services
and Mr. Rob Wilson,  EPA Region 10 for their scrupulous
peer review of this report.
                      DISCLAIMER
This  report has  been  reviewed by  the Office  of Air
Quality Planning  and Standards,  EPA,  and approved for
publication.   Mention  of  trade  names  or  commercial
products  is not  intended to constitute endorsement or
recommendation for use.

-------
                        PREFACE

In this  report  a comparison is made  of  two different
methods  for  estimating the  hourly  Pasquill-Gifford
stability categories required for the current generation
of  regulatory  dispersion   models.    The  effects  of
utilizing  the  two  different  methods  (referred  to  as
Turner   and   SRDT  in  this  report)   in  regulatory
applications of  a Gaussian  dispersion model,  ISC2,  is
also  evaluated.   A fundamental  feature  of  the SRDT
method is the use of on-site meteorological data.

The  Environmental  Protection  Agency must conduct  a
formal and public review before the Agency can recommend
replacement  of   the   Turner  method   for  estimating
stability categories with the  SRDT method.  This report
is being released to establish a basis  for review  of the
consequences  resulting   from  use   of   SRDT-derived
stability categories in routine dispersion modeling of
air pollution impacts.
                          111

-------
                                   CONTENTS


Section                                                                   Page

             Acknowledgements   	  ii

             Preface	iii

             Contents   	iv

             Figures  	   v

             Tables   	vi

   1.         Introduction   	   1

   2.         Rationale	3

   3.         Methods	6

             3.1  Data Selection	6
             3.2  Approach	6

   4.         Stability Comparison Results   	   9

             4.1  Composite Results	9
             4.2  Results for the Kincaid Site	11
             4.3  Results for the Longview Site	14
             4.4  Results for the Bloomington Site	15
             4.5  Results for a AT Interval Other than 2-10m	15
             4.6  Discussion	18

   5.         Results from Dispersion Modeling   	  22

             5.1  Results for the Bloomington Site	22
             5.2  Results for the Kincaid Site	23
             5.3  Results for the Longview Site	23
             5.4  Discussion	23
             5.5  Analysis of Computed Mixing Heights 	  24

   6.         Summary and Conclusions	27

   7.         References   	28

Appendix A   Results of Randomization Analysis  	   A-l

Appendix B   Results from Gaussian Dispersion Modeling  	   B-l
                                      iv

-------
                                    FIGURES


Number                                                                    Page

  4-1      Stability classification plot for composite data using key
          described in Table 4-1; 2-10m AT values (19,540 valid hours)  .  .  13

  4-2      Stability classification plot for Kincaid,  IL data using key
          described in Table 4-1; 2-10m AT (2916 valid hours; 83% of
          the period)	13

  4-3      Stability classification plot for Longview, WA data using key
          described in Table 4-1; 2-10m AT values (8187 valid hours;
          94% of the period)	17

  4-4      Stability classification plot for Bloomington, IN data using
          key described in Table 4-1; 2-10m AT values (8437 valid hours;
          89% of the period)	17

  4-5      Stability classification plot for Kincaid,  IL data using key
          described in Table 4-1 except AT values are from 10-50m (2917
          valid hours; 83% of the period)	20

  4-6      Stability classification plot for Longview, WA data using key
          described in Table 4-1 except AT values are from 10-50m (8187
          valid hours; 94% of the period)	20

  B-l      Mean and median mixing height by hour of the day for
          Bloomington, IN site; 2-10m AT  	   B-7

  B-2      Mean and median mixing height by hour of the day for
          Kincaid, IL site; 2-10m AT	B-8

  B-3      Mean and median mixing height by hour of the day for
          Kincaid, IL site; 10-50m AT	B-9

  B-4      Mean and median mixing height by hour of the day for
          Longview, WA site; 2-10m AT	B-10

  B-5      Mean and median mixing height by hour of the day for
          Longview, WA site; 10-50m AT	B-ll
                                      v

-------
                                    TABLES


Number                                                                    Page

  2-1    -Comparison of incoming solar radiation (insolation)
          classifications 	  4

  2-2     Conceptual matrix for insolation-based key to Pasquill-Gifford
          (P-G) stability categories  	  4

  3-1     Selected on-site meteorological data bases for the SRDT evaluation 7

  4-1     Insolation-based key to Pasquill-Gifford (P-G)  stability
          categories derived from composite data from three sites
          (19,540 valid hours)   	  9

  4-2     Comparison of hourly stability classification via Turner versus
          SRDT for composite data from all three sites using key described
          in Table 4-1; AT values are from 2-10m	10

  4-3     Joint frequency distribution matrix for all SRDT stability
          categories appearing in Table 4-2	10

  4-4     Stability classification results for composite data from all
          three sites using key described in Table 4-1 and AT values
          from 2-10m  (19,540 valid hours)	12

  4-5     Stability classification results for Kincaid, IL data using
          key described in Table 4-1 and AT values from 2-10m (2616
          valid hours; 83% of the period)	12

  4-6     Stability classification results for Longview,  WA data using
          key described in Table 4-1 and AT values from 2-1Om (8187
          valid hours; 94% of the period)	16

  4-7     Stability classification results for Bloomington, IN data
          using key described in Table 4-1 and AT values from 2-10m
          (8437 valid hours; 89% of the period)	16

  4-8     Stability classification results for Kincaid, IL data using
          key described in Table 4-1 except AT values are from 10-50m
          (2917 valid hours; 83% of the period)	19

  4-9     Stability classification results for Longview,  WA data using
          key described in Table 4-1 except AT values are from 10-50m
          (8187 valid hours; 94% of the period)	19

  A-l     Comparison of hourly stability categories via Turner versus
          SRDT for random subsets of the composite data.    . . . V .  . .  .  A-3

  B-l     Design concentration ratios derived from ISC2ST for
          Bloomington, IN site; 2-10m AT   	  B-2

  B-2     Design concentration ratios derived from ISC2ST for Kincaid,
          IL site; 2-10m AT	B-3

  B-3     Design concentration ratios derived from ISC2ST for Kincaid,
          IL site; 10-50m AT	B-4

  B-4     Design concentration ratios derived from ISC2ST for Longview,
          WA site; 2-10m AT	B-5

  B-5     Design concentration ratios derived from ISC2ST for Longview,
          WA site; 10-50m AT	B-6
                                      VI

-------
1.   INTRODUCTION



     The Guideline on Air Quality Models  (Revised)"  (EPA,  1986)  recommends and



ranks four alternative schemes for estimating the Pasquill-Gifford  (P-G)



stability category  (Pasquill, 1961; Gifford, 1961) from on-site meteorological



measurements.  The highest ranking is given to Turner's method  (Turner,  1964)



which uses on-site wind speed coupled with observations of cloud cover and



ceiling height.  However, obtaining the data necessary to implement Turner's



method requires a full time on-site observer, and may be impractical for use  on



a routine basis in many circumstances.



     At the Fourth Conference on Air Quality Modeling, October 1988 (EPA,  1990) ,



public concerns were presented for a practical alternative to the Turner method



for estimating P-G stability categories.  A real need was expressed for  a



method that did not require labor intensive data collection  (e.g., hourly human



observation of clouds), i.e., one based exclusively on simple on-site



meteorological instrumentation.  On February 13, 1991, EPA issued a notice of



proposed rulemaking to further augment the Guideline via Supplement B  (56 FR



5900).  Supplement B  (Draft) included a new method for estimating the



P-G stability category.  In this new method, on-site meteorological



measurements  (10m wind speed in combination with solar radiation during  the day



and temperature difference, AT, at night), are used in lieu of cloud cover and



ceiling height for determining the P-G stability category.  The proposed method



was adapted from Bowen et al.  (1983) and  is herein referred to as the solar



radiation/delta-T (SRDT)  method.  Public  comments presented at the Fifth



Conference on Air Quality Modeling, March 1991  (EPA, 1993) regarding the SRDT



method focused on two key issues:  1) development of the proposed SRDT method



was based on data from only one site  (i.e., Kincaid, IL) for a limited time



period (i.e., 21 weeks during spring/summer); and  2) the method accommodated



AT measurements made only at the 2-lOm interval, whereas AT had been measured



at other intervals by many sources.



     To address these concerns, an attempt was made to acquire several data



bases from diverse geographical areas.  In addition, on-site AT measurements
"Hereinafter,  the  "Guideline"

-------
from other height intervals were considered for evaluation, as available.



Finally, a consequence analysis was needed to document the effect on design



concentration ratios if the new method is implemented.  This report, presented



in seven sections, documents the SRDT evaluation with data from several sites,



and the consequence analysis of using the method in regulatory modeling



applications.  Section 2 of this report presents the rationale behind the



Turner and SRDT methods for determining P-G stability categories.  Section 3 is



a discussion of the methodology used in the analysis.  Section 4 presents and



discusses the results of the stability classification comparison.  Section 5



presents the results from employing the SRDT method in Gaussian dispersion



modeling.  Section 6 provides a summary and conclusions,  and references are



listed in Section 7.  Appendix A contains results from the randomization



procedure used to ascertain the robustness of the SRDT method.  Appendix B



contains tabulated results of design concentration ratios obtained via Gaussian



dispersion modeling.

-------
2.   RATIONALE



     Turbulence, which drives dispersion within the mixed layer of the



atmosphere, is a result of thermal and mechanical processes.  The P-G stability



classification method parameterized these processes using observations  of  wind



speed and subjective estimates of incoming solar radiation  (insolation).



Turner  (1964) provided an objective means for implementing  the P-G method  using



routine airport observations available from the National Weather Service  (NWS).



Stability class, using Turner's method, is a function of wind speed,  the



insolation class (objectively determined based on the sun's position  in the



sky), cloud cover,  and ceiling height.



     Uncertainty in the P-G method arises, in part,  from the subjectivity  in the



classification of insolation.  For example, as indicated in Table 2-1,  Pasquill



(1961) defined strong insolation as:  "... sunny, midday, midsummer conditions



in England."  Based on measurements at Kew Observatory in England, these



conditions correspond to insolation values of about 700 Wm"2, (Chandler, 1965;



Ludwig and Dabberdt, 1972, 1976).  Similarly, Pasquill's definition of  slight



insolation: "... sunny, midday, midwinter conditions in England" corresponds to



insolation values of about 420 Wm"2.




     Insolation flux intensity varies diurnally,  seasonally, and spatially.



There can be significant microscale influences on the amount of insolation



received at the ground surface.  The  intensity and spectral composition of the



insolation are also highly influenced by the amount and type of cloud cover



(Miller, 1981) .  Objective methods for classifying insolation (Table  2-1)



include those of Turner (1964), Ludwig and Dabberdt (1972), Smith (1972) and



Bowen et al. (1983).  Turner's method requires calculation  of the solar



elevation angle based on location and time.  The other methods require  either



estimates or on-site measurements of  insolation.  Strong insolation is  equated



with solar elevations exceeding 60 degrees (Turner,  1964) and insolation values



exceeding 560 Wm'2 (Ludwig and Dabberdt,  1976)  to 700 Wm'2 (Bowen et al.,  1983) .




Slight insolation is equated with solar elevations between  15 and 35  degrees



(Turner, 1964)  and insolation values  less than 280 Wm'2 (Ludwig and Dabberdt,



1976) to 350 Wm'2 (Bowen et al.,  1983).

-------
Table 2-1.   Comparison of  incoming  solar  radiation  (insolation)  classifications.
Source
Pasquill, 1961
Chandler, 1965
Insolation (Wm~2)
Turner, 1964
Solar elevation
(degrees)
Ludwig and Dabberdt, 1972
Insolation (Wrrr2)
Smith, 1972
Insolation (Wm~2)
Bowen et al . , 1983
Insolation (Wm~2)
Strong
sunny , mi dday ,
midsummer
conditions in
England
700"
>60
>560
>600
>700
Moderate Slight Weak
sunny, midday,
midwinter
conditions in
England
420"
35 - 60 15 - 35 <15
280 - 560 <280
300 - 600 <300
350 - 700 <350
" Measurements  made  at  Kew Observatory for conditions  corresponding to
  Pasquill's  definitions.
Table 2-2.  Conceptual matrix  for insolation-based key  to Pasquill-Gifford
             (P-G) stability categories.
DAYTIME
Wind
Speed
(ms-1)

-------
     A desire to retain  the basic  rationale  of  Turner's  method was  an important

 consideration in the  selection of  objective  procedures for use with on-site

 data.  Table 2-2 shows the structural  matrix conceived for the insolation-based

 P-G  stability classification  procedure.   As  explained later,  the specific SRDT

 "cutpoints"  (limits for  on-site meteorological  parameters,  i.e.,  u,  - u«,  E,  -


 E3, ATL/  used to estimate stability categories)  were derived empirically.   The


 SRDT method is  based  on  a development  by Bowen  et  al.  (1983),  with

 modifications as necessary to retain as  much as possible of the structure and

 behavior of Turner's  method as implemented in the  EPA meteorological

 preprocessors for  regulatory  models  (EPA,  1986).   The first modification  was to

 replace  Bowen's method for determining nighttime (insolation less than 35 Wm'2)


 with the procedure which is based  on calculations  of sunrise and sunset.   This

 modification was necessary to maintain consistency in the SRDT method and that

 used in  EPA's meteorological  preprocessors.   Another modification was to

 include  an additional daytime insolation class  (
-------
3.   METHODS

3.1  Data Selection

     A search produced 10 potentially suitable data bases.  Of these, 3 were

selected as best meeting the requirements for this evaluation  (Table  3-1).  The

requirements for individual data bases  included the following attributes:  1)

hourly average values for 2-10m temperature difference," 10m wind speed and

direction, and total solar radiation; 2) available cloud  cover and  ceiling

height data from a nearby, representative NWS station;  3) a continuous

monitoring record of sufficient length  (preferably, at  least one  full year);

4) on-site meteorological monitoring having been done in  accordance with EPA

guidance  (EPA, 1987a); and 5) on-site meteorological data having  been quality

assured.  There was also a desire to acquire data bases that, in  the  aggregate,

were geographically within the contiguous United States.

3.2  Approach


     As mentioned above,  the method with which to compare the SRDT  system is

that prescribed by Turner (1964), as implemented in the Meteorological

Processor for Regulatory Models (MPRM)(Irwin et al., 1988), hereafter referred

to as the "Turner method".  The Turner method uses on-site wind speed coupled

with cloud cover and ceiling height observed on-site.   Because on-site data for

cloud cover were unavailable, surface observations from a nearby, represen-

tative NWS station were used as surrogates.  Accordingly, on-site data bases

were carefully selected (see Section 3.1) to ensure the integrity of  their use

with surrogate NWS data.  Determination of P-G stability  categories was made

using MPRM (Version 1.3),  configured to implement the Turner method.

Consistent procedures were used for all sites.  For stabilities determined

using either the Turner method  (via MPRM6)  or  the  SRDT method,  "smoothing"

(i.e., disallowing stability to change by more than one class per hour) was

disabled in this evaluation in an effort to make a direct comparison  of the


stability categories generated by both methods.  In all evaluations,  quality

control measures were implemented to ensure that only data valid for  joint
"Other  intervals  were  also  of  interest.

bA special  version of  MPRM  1.3 (MPRMRUFF)  was configured to output "rough"
hourly stability classes.

-------
Table 3-1.  Selected on-site meteorological data bases for the SRDT evaluation.
                                                                                      AT Height Interval  (m)
  Source    Location
NWS Station
Distance"
Period     Insolation    2-10
                                                                                                   Other
  EPRI"     Kincaid, IL
Springfield,  IL
   -25     4/80  -  8/80
               Yes
Yes    10-50, 10-100
  ENSR°     Longview, WA
Portland,  OR
   -55     1/91 - 12/91
               Yes
Yes
                                     2-50
  ENSR     Bloomington, IN    Indianapolis, IN
                      -70
            7/91  -  7/92
               Yes
Yes
"Kilometers  from nearest  NWS  station.
""Electric  Power Research  Institute;  data base  used in original SRDT evaluation by EPA (see Section 1.0) .
"ENSR,  Inc.;  data were  collected at  a pulp and paper  mill  operated by Weyerhauser,  Inc.

-------
 stability comparison were used.   All other data were bypassed but otherwise
 accounted for (in all comparisons,  valid data ranged from 83 to 94 percent).
      Once  the requisite  P-G stability categories were determined via MPRM for
 each site,  the SRDT system was applied.   The SRDT system uses on-site wind
 speed,  total solar radiation (daytime)  and temperature difference (AT)
 (nighttime).  The temperature differences were measured with reliable
 thermocouple systems.  Outpoints for the solar radiation and AT parameters were
 derived iteratively to obtain optimal fits for the entire time period at each
 site.   The  observed range of direct solar radiation intensities reported for
 contiguous  U.S.  locations (Miller,  1981)  was investigated in an effort to
 develop a  daytime scale  that would  be geographically robust.   Initial
 evaluations indicated some site-to-site  variations in the derived cutpoints.
 Therefore,  it was decided to pool the data from all three sites to determine
 cutpoints  from the composite data set.   The occurrence of residuals  (category
 differences)  on  an hourly basis  was minimized;  attention was  paid to the
 distribution of  those  residuals  by  category and a systematic  effort  was
 employed in the  choice of cutpoints to evenly allocate those  residuals  across
 all  stability categories  in an attempt to make  the system as  robust  as
 possible.   These  cutpoints were  then applied to each individual data base to
 assess  site-specific residuals in the behavior  of the SRDT method.
     As a further effort  to  investigate the  sensitivity of the  results,  the
 composite data were randomly stratified  into two complementary (mutually
 exclusive)  subsets; hourly records  for which information  was  valid for  joint
 stability classification  were randomly sorted into two bins.   Records from each
 bin were then used independently to  evaluate  the  SRDT method.   Cutpoints
 determined for the pooled data were  applied  individually  to each bin.   This
 approach allowed for an assessment of the  sensitivity of  the  SRDT  results  to
 the specific data employed in the analyses.   Thus,  for the composite data  base,
 results could be assessed as random  fluctuations  over  many iterations.
     Finally, a consequence analysis showing effects on design concentration
ratios was performed using three hypothetical sources, a hypothetical receptor
array on flat terrain, and a suitable Gaussian dispersion model  (ISC2;  see
Section 5).

-------
4.   STABILITY COMPARISON RESULTS

4.1  Composite Results

     For the pooled analysis from the three data bases  (i.e., Kincaid, Longview,

and Bloomington),  19,540 hours  (89.6% of  those potentially available  for 909

days) were valid for making the joint comparison of  stability classes.   As

indicated in Table 4-1, the optimum  outpoints for  solar radiation were 925,

675, and 175 Wm"2.  Daytime wind speed cutpoints were 2.0,  3.0, 5.0, and  6.0 ms"1;

those for nighttime wind speed were  2.0 and 2.5 ms"1.  Using these  cutpoints,

comparison of hourly stability categories for both methods showed reasonable

agreement (Table 4-2).  The joint frequency distribution of hourly stability

categories modulated via the SRDT method was examined  (Table  4-3).  Of most

interest was the discrimination made at night as a function of wind speed and

AT.  Most of the category "sorting skill" is being made on the basis  of  wind

speed, with AT adding a refinement.  The weak discrimination  seen with

nighttime AT has been observed by others  (Bowen et al.,  1983; Bowen and  Pamp,

1994).  To check this phenomenon, the nominal value  for the AT cutpoint  (ATL)

was varied iteratively from 0.0, -0.01, -0.02, -0.03, +0.01,  +0.02, and  +0.03.

No systematic improvement was seen over that using ATL = 0.0,  the value employed
Table 4-1.   Insolation-based key to Pasquill-Gifford (P-G)  stability categories
             based on  composite  data from three sites.

Wind
Speed"
(ms-1)
<2.0
2.0 -* 3.0
3.0 -» 5.0
5.0 -> 6.0
>6.0
DAYTIME"
Solar Radiation (Wm"2)
>925
A
A
B
C
C
925 -» 675
A
B
B
C
D
675 -» 175 <175
B
C
C
D
D
D
D
D
D
D
NIGHTTIME"
Wind
(ms-1)
<2.0
2.0 -» 2.5
^2.5
2 -10m AT (°Cm-')

-------
Table 4-2.   Comparison of hourly stability classification via Turner versus
             SRDT for composite data from all three sites using key described
             in Table 4-1;  AT values are from 2-10m.
SRDT

A
B
C
Dday
Anight
E
F
TOTAL
%"
P-G Stability Categories as Estimated via Turner
A B C D^ D^ E F&G
108
118
31
4
0
0
0
261
41.4
160
1230
429
244
0
0
0
2063
59.6
3
393
970
1037
0
0
0
2403
40.4
2
252
996
3787
0
0
0
5037
75.2
0
0
0
0
2085
659
892
3636
57.3
0
0
0
0
905
250
55
1210
20.7
0
0
0
0
651
662
3617
4930
TOTAL
273
1993
2426
5072
3641
1571
4564
19540
73.4
  Percent coincidence of the hourly stability categories based on the
  distribution derived via Turner (see Section 3.2).
Table 4-3.   Joint  frequency distribution matrix  for  all  SRDT stability
             categories appearing  in  Table  4-2.b
ws
<2.0
2.0-3.0
3.0-5.0
5.0 - 6.0
>6.0
TOTAL
SOLAR RADIATION (Wnf2)
>925
25
50
72
12
3
162
925-675
198
341
493
114
86
1232
675-175 <175
1087
826
1471
401
349
4134
1676
925
1156
246
233
4236
AT/AZ (°Cm-1)
WS
<0.0 2:0.0

<2.0
2.0 - 2.5
>2.5


549
188
773

1510
4564
1022
2680



8266 I 19540
b The  composite  data valid for comparison comprised 9764  daytime hours and
  9776 nighttime hours.
                                      10

-------
in the SRDT method examined by Bowen et al.  (1983).  It was therefore decided



to retain the cutpoint at 0.00Cm"'.  The nighttime wind speed cutpoints were



likewise varied  iteratively over reasonable values.  The proximity of the



chosen cutpoints (0.5 ms'1 apart)  was necessary to get the required sorting



skill in concert with AT, and no alternative AT cutpoint  (ATL)  allowed the two



nighttime wind speed cutpoints to be any further apart than 0.5 ms"1.



     Overall,  the stability classifications for the two methods coincided for



62% of the hours, and were within one category for 89% of the hours  (Table



4-4; Fig. 4-1).  Absolute residuals  (|A|)  expressed  as a  percentage  of hours



allocated to each category, were analyzed by stability category, and by day



versus night.  Across all categories, the mean residual was less than one



percent.  The mean absolute residual was greater for nighttime hours than for



daytime hours.  As indicated in Table 4-2, however, the coincidence of



categories by both methods varied as a function of stability category.  The



greatest coincidence (75%) occurred with daytime D's, while the least (21%)



occurred with E's.   As these results were considered to be optimum for the



pooled data, the cutpoints were applied to the individual sites to assess their



residuals.



     The results of the randomization analysis are presented in Appendix A.  The



SRDT method was not seen to be sensitive to random variations in the data.   The



results for complementary subsets of the pooled data were virtually identical.



4.2  Results for the Kincaid Site



     The first data base examined" was from the Electric Power Research



Institute (EPRI) Plume Model Validation and Development Program  (PMVDP)  for  the



plains site, Kincaid, Illinois.  The meteorological monitoring site is located



in central Illinois; the surrounding terrain is flat and uniform  (z0 - 10cm).



The site and its environs have been extensively described elsewhere  (EPRI,



1983).   Though meteorological measurements were made from March through



November, 1980, data for a 21-week period  (7 April - 31 August 1980) were



considered to be of highest quality.  On-site measurements of interest included



those of 10m wind speed and direction, 2-10m, 10-50m and 10-100m AT, and total
"These  data were  used in  the  analysis  for EPA's  initial  proposal to adopt SRDT.



                                       11

-------
Table 4-4.   Stability classification results  for composite data from all three
             sites  using key described in Table 4-1  and AT values from 2-10m
             (19,540  valid hours).
STABILITY
A
B
C
D*,

Anight
E
F


Turner (%)
261 (1.3)
2063 (10.6)
2403 (12.3)
5037 (25.8)

3636 (18.6)
1210 (6.2)
4930 (25.2)


SRDT (%)
273 (1.4)
1993 (10.2)
2426 (12.4)
5072 (26.0)

3641 (18.6)
1571 (8.0)
4564 (23.4)


|A| (%)
(0.1)
(0.4)
(0.1)
(0.2)
^
(0.0)
(1.8)
(1.8)
-

Mean (%)




(0.193)



(1.20)
(0.62)
Hourly coincidence of stability categories:  61.7%
Hourly categories ± one class:               89.4%
Table 4-5.   Stability  classification results  for Kincaid,  IL  data  using key
             described  in  Table 4-1  and  AT values from  2-10m (2916  valid hours;
             83% of  the period).
STABILITY
A
B
C
Dday

Anight
E
F


Turner (%)
61 (2.1)
376 (12.9)
497 (17.0)
618 (21.2)

322 (11.0)
237 (8.1)
805 (27.6)


SRDT (%)
42 (1.4)
301 (10.3)
432 (14.8)
777 126.6)

611 (21.0)
257 (8.8)
496 (17.0)


|A| (%)
(0.7)
(2.6)
(2.2)
(5.4)
-
(10.0)
(0.7)
(10.6)
-

Mean (%)



- -
(2.7)



(7.1)
(4.6)
Hourly coincidence of stability categories:  56.4%
Hourly categories ± one class:                89.7%
                                       12

-------
   3D
(J
C
O>
U
U
o
 Figure 4-1.
     B           C          D        DCnite}       E

             P-G  Stabi I  i ty  Category


Stability classification plot for composite data using key
described in Table 4-1; 2-10m AT values (19,540 valid hours)
   3D
                          P-G  Stabi  I 1ty  Category
Figure 4-2.  Stability classification plot  for Kincaid, IL data using  key
             described in Table 4-1; 2-10m  AT values  (2916 valid hours;
             83% of the period).
                                      13

-------
 solar radiation.   In addition to the on-site data,  concurrent surface obser-



 vations  from the  NWS station at Springfield,  IL (WBAN #93822),  about 25km



 northwest  of the  site,  were  obtained.



      For the  Kincaid data base,  2616 hours  (83% of  those potentially available



 for  147  days)  were valid for making the  joint comparison of stability classes.



 Overall, the stability  classifications for  the two  methods coincided for 56% of



 the  hours, and were within one category  for 90% of  the hours (Table 4-5; Fig.



 4-2).  Unstable and E categories were comparable.   The SRDT method



 overrepresented the nighttime D category and underrepresented the F category.



 4.3  Results  for  the  Longview Site



     ENSR Consulting  and Engineering provided the next  data base.   The



 meteorological monitoring site is located in Longview,  in  southwest Washington,



 along  the Columbia River approximately 65km inland  from the Pacific Ocean.



 While  the terrain within the  local  proximity of the city of Longview is



 relatively flat (z0 - 10cm),  the terrain immediately across the Columbia River



 in Oregon and just outside the Longview  city limits in Washington ascends



 quickly into  a series of ridges  and hills.   The city itself (and  the monitoring



 site)  is approximately  5m above  mean sea  level  (msl)  due to its low-lying



 position along the Columbia River.  Terrain extends 60m above msl within 3km of



 the monitoring site.  Data for calendar year  1991 were  available  for this



 analysis and  were  collected and  quality assured according  to EPA  guidance  (EPA,



 1987a), as well as ENSR's own internal standard operating  procedures.  On-site



 measurements  of interest included those of  10m wind speed  and direction, 2-10m



 and 2-50m AT, and  total solar radiation.  In  addition  to the on-site data,



 surface observations  from the  NWS station at  Portland,  OR  (WBAN #24229), about



 65km southeast of  the site,  were  obtained.  The topographical setting for the



 site, unique among  the three  sites  examined in this  analysis, is  such that



 local micrometeorologial effects  are possible.  Preliminary analyses of



nighttime wind speeds indicated  that the site is influenced by nighttime



drainage flows, resulting in  relatively low (33% si ms"1; 62% s2 ms'1;



 u  =1.9  ms"1)  and  uniform velocities.




     For  the  Longview data base, 8187 hours  (94% of those  potentially available



for 365 days) were valid for making the joint comparison of  stability classes.





                                       14

-------
Overall, the stability classifications for the two methods coincided for 58% of

the hours, and were within one category for 85% of the hours  (Table 4-6; Fig.

4-3).   Unstable categories generally compared better, while  (as with Kincaid)

some disparity occurred for the nighttime D category.

4.4  Results for the Bloomington Site

     ENSR Consulting and Engineering also provided the third data base.   The

meteorological monitoring site, equipped with a 10m tower, is located in a

rural area with slightly rough terrain (z0 - 25cm)  about 70km south of

Indianapolis, near Bloomington, IN.  To compensate for several days of missing

data due to frequent lightning-caused outages, data for the  13-month period

July 1991 - July 1992 were provided.  These data were collected and quality

assured according to the provisions of EPA guidance  (EPA, 1987b), and ENSR's

own internal standard operating procedures.  On-site measurements of interest

included those of 10m wind speed and direction, 2-10m AT, and total solar

radiation.  In addition to the on-site data, concurrent surface observations

from the NWS station at Indianapolis, IN  (WBAN #93819) were  obtained.

     For the Bloomington data base, 8437 hours (89% of those potentially

available for 397 days) were valid for making the joint comparison of stability

classes.  Overall, the stability classifications for the two methods coincided

for 67% of the hours, and were within one category for 94% of the hours  (Table

4-7; Fig. 4-4).  As with Kincaid, but to a  lesser extent, the SRDT method

underrepresented the F stability category.

4.5  Results for a AT Interval Other than 2-10m

     There was interest to investigate the performance of the SRDT method for AT

values measured at intervals above 10m.  On-site meteorological data  collected

at Kincaid and Longview afforded just such  an opportunity, as they included  10-

50m AT values.  As a group, the 10-50m AT values at both sites were less

variable  in absolute magnitude and less frequently in the isothermal/positive

range as  compared with their 2-10m counterparts.'  Therefore, it was

anticipated that a AT cutpoint of slightly  less than 0.0 "Cm'1  (e.g.,  -0.01 or
"For Kincaid,  95% of the 2-10m AT values were in the isothermal/positive range,
versus 79% of those  from 10-50m.  Likewise,  for Longview,  89%  of  the  2-10m AT
values were in the  isothermal/positive  range, versus 46%  of  those from  10-50m.


                                        15

-------
Table 4-6.  Stability classification results for Longview, WA data using key
            described in Table 4-1 and AT values from 2-10m (8187 valid hours;
            94% of the period).
STABILITY
A
B
C
Dday

Dnigfc
E
F


Turner (%)
81 (1.0)
880 (10.7)
829 (10.1)
2153 (26.3)

1682 (20.6)
478 (5.8)
2084 (25.4)


SRDT {%)
175 (2.1)
993 (12.1)
776 (9.5)
1999 (24.4)

1268 (15.5)
714 (8.7)
2262 (27.6)


|A| (%)
(1.2)
(1.4)
(0.6)
(1.9)
— »
(5.0)
(2.9)
(2.2)
-

Mean (%)




(1.3)



(3.4)
(2.2)
Hourly coincidence of stability categories:     58.0%
Hourly categories ± one class:                  85.0%
Table 4-7.  Stability classification results for Bloomington, IN data using key
            described in Table 4-1 and AT values from 2-10m  (8437 valid hours;
            89% of the period).
STABILITY
A
B
C
D*y

Dnieto
E
F


Turner (%)
119 (1.4)
807 (9.6)
1077 (12.8)
2266 (26.9)

1632 (19.3)
495 (5.9)
2041 (24.2)


SRDT (%)
56 (0.7)
699 (8.3)
1218 (14.4)
2296 (27.2)

1762 (20.9)
600 (7.1)
1806 (21.4)


|A| (%)
(0.7)
(1.3)
(1.6)
(0.3)
-
(1-6)
(1.2)
(2.8)
-*

Mean (%)



-
(1.0)



(1.9)
(1.4)
Hourly coincidence of stability categories:     66.9%
Hourly categories ± one class:                  93.5%
                                       16

-------
  30
                          P-G  Stabi I  ity  Category
Figure 4-3.  Stability classification plot for Longview,  WA data using key
             described in Table 4-1; 2-10m AT values (8187 valid hours;
             94% of the period).
                          P-G  Stab I I  ity  Category

Figure 4-4.  Stability classification plot for Bloomington, IN data using key
             described in Table 4-1; 2-10m AT values (8437 valid hours;
             89% of the period).
                                      17

-------
-0.02) would have shown greater skill for the 10-50m AT's.  Iteratively,  it was



found that a value of -0.01 "Cm"1 seemed to produce only marginally better




results than those using 0.0 °Cm''.  For Kincaid, the nighttime mean absolute




residual  (|A|) was  6.9% for ATL = -0.01 and -0.02  (versus 7.9% for ATL = 0.0).




For Longview,  |A| was 3.9% for ATL = -0.01 and  -0.02 (versus 4.3% for ATL =




0.0).  Temperature difference  offers only a minor refinement to the



determination of stability category. Therefore, it was decided to employ  a



0.0 "Cm"1 cutpoint,  regardless of measurement height interval.




     The results for Kincaid (Table 4-8; Fig.  4-5)  were similar to those using



the 2-10m AT's; the mean residual for nighttime stability categories was  about



7 percent.  For the Longview site (Table 4-9; Fig. 4-6), the mean residual for



nighttime stability categories was about 4 percent, slightly greater  (4.3%)



than that using the 2-10m AT's  (3.4%).  The analyses for both sites serve to



show that stability categories can be as reasonably determined using AT values



measured above 10m using the same AT criteria as for 2-10m.  Indeed, analyses



with more data bases may better support the notion that use of a AT cutpoint



somewhat less that 0.0 °Cm'! affords better classification skill for intervals




above 10m.  However, for practicality in implementing the SRDT method it  was



considered that 0.0 "Cm"1 was reasonable, particularly given site-to-site




variability seen among the data bases used here.



4.6  Discussion



     Of interest in illustrating how the mechanics of the SRDT method work, it



may be noted that the D stability category at night occurred for about 12 to  15



percent fewer hours at the Longview site than those examined at the Bloomington



and Kincaid site, respectively.  This difference is primarily explainable by



the nighttime wind regime.  At the Longview site,  nighttime winds s2 ms~'  (a



requirement for stable classification for any lapse rate) occurred 11 to  25



percent more frequently than did those at Bloomington and Kincaid, respec-



tively.  It is thought that such a regime at the Longview site is due to



micrometeorological effects (see Section 4.3).



     Also,  as indicated by data from the Kincaid and Longview sites, prevalence



of nighttime D categories increased for the 10-50m AT data compared to their
                                       18

-------
Table 4-8.   Stability  classification results  for Kincaid,  XL  data using key
             described  in Table  4-1  except  AT  values  are  from  10-50m (2917  valid
             hours;  83% of  the period).
STABILITY
A
B
C
D*y

Dnigh.
E
F


Turner (%)
61 (2.1)
376 (12.9)
497 (17.0)
618 (21.2)

323 (11.1)
237 (8.1)
805 (27.6)


SRDT (%)
42 (1.4)
301 (10.3)
432 (14.8)
777 (26.6)

631 (21.6)
277 (9.5)
457 (15.7)


|A| (%)
(0.7)
(2.6)
(2.2)
(5.4)
-
(10.5)
(1.4)
(11.9)
-

Mean (%)




(2.7)



(7.9)
(5.0)
Hourly coincidence of stability categories:     56.1%
Hourly categories ± one class:                  90.5%
Table 4-9.  Stability classification results for Longview, WA data using key
            described in Table 4-1 except AT values are from 10-50m  (8187 valid
            hours; 94% of the period).
STABILITY
A
B
C
EV

Anight
E
F


Turner (%)
81 (1.0)
880 (10.7)
829 (10.1)
2153 (26.3)

1682 (20.5)
478 (5.8)
2084 (25.4)


SRDT (%)
175 (2.1)
993 (12.1)
776 (9.5)
1999 .(24.4)

1553 (19.0)
1011 (12.3)
1680 (20.5)


|A| (%)
(1.2)
(1.4)
(0.6)
(1.9)
-
(1.5)
(6.5)
(4.9)
-

Mean (%)



- _
(1.3)



(4.3)
(2.6)
Hourly coincidence of stability categories:     56.6%
Hourly categories ± one class:                  88.8%
                                       19

-------
(D
U
CZ

OJ
13
(J
U
o
   3D
   25  -
   2O  —
   15  —
   10  -
    5  -
                           P-G  Stab I I  ity  Category



 Figure 4-5.   Stability classification plot for Kincaid, IL data using key

              described in Table 4-1 except AT values are from 10-50m (2917

              valid hours; 83% of the period).
O)
U
b
U
U
o
   30
   25  -
   2O  —
   10  -
    5  -
                           P-G  Stabi I  ity  Category
        A
 Figure 4-6.  Stability classification plot for Longview, WA data using key

              described in Table 4-1 except AT values are from 10-50m (8187

              valid hours; 94% of the period).
                                       20

-------
2-10m AT counterparts at both  sites.  The  increase is due  to  the  less  frequent
occurrence of isothermal/positive  lapse rates seen in the  upper boundary  layer
 (Section 4.5), and the increase seen is roughly proportional  to the  decrease  in
occurrence of such lapse rates.
     For the analysis of stability comparisons,  Tables 4-4 through 4-9 emphasize
the overall comparability of the frequency of occurrence of stability
categories in the aggregate, i.e., without regard to hourly correspondence.   As
explained in Section 3.2, a systematic effort was employed to evenly allocate
residuals on an hourly basis across all stability categories.  Detailed hourly
correspondence of stability categories was analyzed for all comparisons but
reported (in matrix format) only in Table  4-2 for the pooled  data.   For 2-10m
AT measurements, hourly correspondence of  stability categories ranged  from 56
to 58 percent for three data bases analyzed; categories were  within  one class
for 85 to 94 percent of the hours examined.  For 10-50m AT measurements, hourly
correspondence of stability categories ranged from 56 to 57 percent  for two
data bases analyzed; categories were within one class for  89  to 91 percent of
the hours examined.  As indicated in the matrix for the pooled data  (Table 4-2),
infrequently the corresponding categories  differed by two  classes or more.  An
important point to remember in viewing these results is that  having  a  stability
category that differs by no more than one  class most of the time on  an hourly
basis can still result in quite different  design concentrations as different
wind speeds, directions and mixing heights are being linked with those
stability categories.
     These  results suggest use of a nighttime AT cutpoint value of 0.0 ''Cm'1  is
robust enough to accommodate a range of height intervals, provided attention  is
given to proper siting of temperature sensors so as to effectively characterize
the boundary layer.  Consistent with probe placement guidance (EPA,  1987a),  the
lower temperature sensor should be least in the range of order of 20z0  -  100z0,
but never less than 1m,  above the ground surface (Irwin et al.,  1985) .  The
upper sensor should be of the order 5 times the lower sensor height.   These
criteria ensure that the lower instrument  level is within the surface  layer and
that a reasonable separation is maintained between the two measurement levels.
Stronger temperature gradients are expected in the lower atmosphere.   Hence,   as
the distance above ground increases for the lower measurement level,  the
separation  distance accordingly should increase.
                                       21

-------
5.   RESULTS FROM DISPERSION MODELING

     To ascertain the possible effect of the new SRDT stability classification

method on design concentrations, a sensitivity analysis was performed on the

ratios of design concentrations  (XSRDr/XTuror) .  The  Industrial  Source Complex

(ISC2) model was used to compute concentration values for  averaging times of

1-hour, 3-hour, 24-hour, and the entire period modeled.  ISC2 provides both the

high first high  (H1H) and the high second high  (H2H) concentrations.   Three

stationary point sources" of heights  35m,  100m,  and 200m,  respectively, were

used in these analyses.  These same sources have been used in past  modeling

evaluations to assess the impact of revisions to regulatory air quality models

(Lee et al., 1979).  Receptors were arranged in a  polar grid network with 36

radials and 180 sites on 5  concentric rings at 800m, 2000m, 4000m,  7000m, and

15000m, respectively; the sources were placed at the origin.  Flat  terrain was

assumed and the model was run in the RURAL mode.   Hours with  on-site wind speed

less than 0.5 ms"1 were treated as calms, and the option 'MSGPRO', which allows

processing of missing hours,b was set 'ON'.   For analyses of 24-hour concen-

trations, days with more than 6 missing hours were omitted.  The  daytime

morning mixing height is determined daily within the meteorological processor

based on the stability category just before sunrise.  The  maximum afternoon

mixing height  (ziKnax)  was  preset  to  2500m.   This  configuration conferred a

measure of consistency between the ISC2 runs.  The tabulated  results for all

ISC2 runs are presented in  Appendix B.

5.1  Results for the Bloomington Site

     At the Bloomington, IN site, where the AT interval was 2-10m,  the composite

mean ratio  (across 4 averaging times, 3 source heights, and both  concentration

types; 24 values) was 1.06  (median was also 1.06), with a  range  (R)  of 0.85 -

1.24  (Table B-l); the geometric mean ratio was 1.05.  For  the H1H concen-

trations  (12 values), the mean ratio was 1.07  (R = 0.97  -  1.16),  with a median

of 1.05; the geometric mean was 1.07.  For the H2H concentrations (12 values),
"For  the  35m stack,  parameters were:  Q8 = 100 gs"1, T, = 432K,  v. = 11.7 ms"1  d,  =  2.4m
For the 100m stack, parameters were: Q. = 100 gs'1,  T.  = 416K, v, = 18.8 ms ,  d, = 4.6m
For the 200m stack, parameters were: Q, = 100 gs"1,  T,  = 425K, v, = 26.5 ms"1,  d, = 5.6m

bSuch hours are also processed as calms.

                                       22

-------
the mean ratio was 1.05  (R = 0.85 - 1.24), with a median of 1.07; the geometric
mean was 1.04.
5.2  Results for the Kincaid Site
     For analyses done with 2-10m AT data, the composite mean ratio was 1.06
(median was 1.08), with a range of 0.75  - 1.62  (Table B-2); the geometric mean
ratio was 1.05.  For the H1H concentrations, the mean ratio was 1.09  (R = 0.77
- 1.62), with a median of 1.12; the geometric mean was 1.08.  For the H2H
concentrations, the mean ratio was 1.04  (R = 0.75 - 1.41), with a median of
1.05; the geometric mean was 1.01.
5.3  Results for the Longview Site
     For analyses done with 2-10m AT data, the composite mean ratio was 1.24
(median was 1.20), with a range of 1.00  - 1.70  (Table B-4); the geometric mean
ratio was 1.22.  For the H1H concentrations, the mean ratio was 1.25  (R =
1.00 - 1.70), with a median of 1.20; the geometric mean was 1.23.  For the H2H
concentrations, the mean ratio was 1.24  (R = 1.00 - 1.62), with a median of
1.19; the geometric mean was 1.22.
5.4  Discussion
     As noted in the above results (as well as in the footnotes for Tables B-l
to B-5), for each site and'AT interval a composite mean ratio was computed
based on 24 values.  Mean ratios were also computed for the H1H and H2H
concentrations based on 12 values for each.  Likewise, the geometric mean and
standard deviation were also computed and reported.  Because the design
concentration ratios are considered to be approximately log-normally
distributed, the latter statistics probably better characterize the ratio
distribution.  Formal hypothesis testing is impractical as strict independence
among the ratios cannot be assumed.   For example, a concentration-value that
figures into a ratio for one averaging time may also figure into a ratio for a
longer averaging time at the same site.
     The ratios at both the  Kincaid  and Bloomington sites do not appear to be
significantly different than 1.0.  Nor is there sufficient evidence to warrant
the conclusion that design concentrations predicted via SRDT-derived
stabilities are from a different population than those predicted via Turner-
derived stabilities.  While the same relationship does not appear to be the
case with the Longview data,  neither can meaningful confidence bounds be

                                       23

-------
established.  As explained in Section 4.6, there may be some site-specific
features at the Longview site that  result in some unusual influences.  Taken
on the whole, the range of concentration differences seen among the three sites
is of the order expected for site to site differences and professional
judgement should be used in viewing the modeling results presented.
5.5  Analysis of Computed Mixing Heights
     For ratios 21.15, special attention was given to the influence of computed
mixing height values for the averaging times involved.  Such values are
themselves determined by the occurrence of stability category, and can be
highly influential in the model predicted concentrations.  Thus, values for
estimated mixing heights in the short term, as well as their long term
distribution pattern, was of interest in assessing the consequence of the SRDT
method on concentration values; one must interpret such concentrations in the
context of the associated mixing heights.  The behavior of the computed mixing
heights was investigated using simple statistics, and the diurnal patterns are
depicted in Figures B-l to B-5.  In these Figures, the mean  (zt) (Figs. B-la to
B-5a) and median (Figs. B-lb to B-5b)  mixing height was determined by hour of
the day.  For convenience, only daytime hours were analyzed.
     At the Bloomington site,  on three occasions the ratio was al.15.   The
associated mixing heights were not seen to have been influential in the
prediction of the higher ambient concentration via SRDT-derived stability
categories.  While the period averages were consistently higher via SRDT, the
period averaged daytime mixing heights via Turner (2150m; median = 2500m) were
lower than those computed via SRDT (2170m; median = 2500m).  Therefore, mean
mixing height does not adequately explain the high concentrations for the
period.
     At the Kincaid site,  on 8 occasions the ratio was si.15.  The associated
mixing heights were seen to be increasing the ratio in half of these instances,
though the pattern appeared to be random.  The period averages were higher via
SRDT only for the 35m stack.  As with the Bloomington site, the period averaged
daytime mixing heights via Turner (2040m; median = 2500m) were lower than those
computed via SRDT (2130m;  median = 2500m).  When stability category was
estimated using the 10-50m AT data,  the results were virtually identical to
those found with the 2-10m AT data (compare Tables B-2 and B-3).

                                       24

-------
     At the Longview site, on 14 occasions  the  ratio  was  21.15.   In some of
these cases, differences  in  early morning daytime mixing  heights seemed to
account for the differences  seen in  concentration values.  This was especially
true for the 200m source.  As with the  Kincaid  site,  period averages were
consistently higher via SRDT.   However,  the period averaged daytime mixing
heights via Turner  (2150m; median =  2500m)  were greater than those computed via
SRDT (2130m; median = 2500m).   When  stability category was estimated using the
10-50m AT data, the results  were virtually identical  to those found with the
2-10m AT data  (compare Tables B-4 and B-5).   The period averaged daytime mixing
heights via Turner using  the 10-50m  AT  data (2470m; median = 2500m) were also
greater than those computed  via SRDT (2150m;  median = 2500m).   For this site,
the mean mixing height may at least  partially explain the high concentrations
for the period.
     In general, at all sites and for both  AT intervals,  the mean mixing height
at or just after sunrise  computed by MPRM is  greater  with SRDT-derived
stabilities  (z^^j.)  than with those  derived via Turner (z,_ri(mCT.) stabilities
(Figs. B-la to B-5a) .  Whereas, except  for  at Kincaid, mean afternoon"  z^^j.
seems to be lower than  zi_Ttmier .
     This disparate behavior in the  time  series of  computed mixing heights is an
inherent trait of the mixing height  algorithm as implemented in MPRM and can,
in part, be traced directly  to  how mixing heights are determined for daytime
hours having neutral stability.  In  these cases,  the  algorithm interpolates in
time by one of two different algorithms,  depending on the estimated stability
just before sunrise.  This may  result in  spurious increases and decreases in
the time series of early morning daytime  mixing height values.
     At the Bloomington site, the early morning ~z,_SKDT  is  considerably higher
than zi_Txnier .  This is mainly  due to  the  greater number of  nighttime hours with D
stability via SRDT.   In mid-morning  this  pattern reverses,  followed by
convergence at about 1400 hours.  At the  Kincaid site, the early morning z,_S{DT
is also higher than zl_Tlimer, again due to a greater number of nighttime hours
having D stability via SRDT.  After  a mid-morning convergence, z^^^ again
exceeds z{_Tltnifr,  followed by convergence  at about 1400  hours.  Note  that  'Z^^T is
"I.e.,  from about 0900 -  1400 hours.
                                       25

-------
 never lower than z(_Ttma. .   At Longview, the pattern is not quite  so predictable/
 explainable as that for  the other sites.   For the 2-10m AT data, the early
 morning  z,.Tmtr is initially higher than z.-.^oj-, due to a greater number of
 nighttime  hours having D stability via Turner.   By 0600 hours, though, the
 pattern  reverses.   Following a mid-morning convergence, there is another
 reversal,  with zt.Tmfr again higher than ~zi.SRDT; by 1400 hours the  z,.'s converge.
 For  the  10-50m AT data,  the pattern is the same as that for the 2-10m AT data
 except that the early morning ~zi.stDT  is higher than ~z,_Txmer, even though there
 were still 8 percent more nighttime hours with  D stability via Turner!  Perhaps
 some of  the aforementioned micrometeorological  influences (Section 4.6)  were
 operating  in some more complex way for the computation of mixing heights at
 this site.
     The occurrence  of a  larger early  morning mixing height  is largely related
 to the prevalence of nighttime D  stability category  (specifically,  its
 occurrence the hour before sunrise).   This relationship was  borne out in all
 mixing height analyses except  for that using the 10-50m AT data at the Longview
 site (Figure B-5a),  where the  relationship was  reversed.   The mixing height
 analyses serve to illustrate how  the prediction of ambient  concentrations is
 affected by mixing  heights', which themselves exhibit complex patterns due to
 the  influence of stability categories  and their occurrence  relative  to the time
 of sunrise.   Because of the nonlinear  linkage between  the occurrence of
 stability  category  and predicted  concentration  via factors  such as mixing
 height, care should be exercised  in  interpreting the stability comparison
 results and those from the dispersion  modeling  (see discussion at end of
 Section 4.6).
     Thus,  in the comparisons made in  this evaluation,  apparent  disparities  in
mixing heights, which may indeed  result in the  prediction of significant
 concentration values,  are  as likely as not  an artifact  of the computational
 system.  Though it was possible,  it was not  deemed prudent to "factor"  the
mixing height  influence out because it would not have  emulated the complete
computational  system as it is  employed for making model predictions  in
regulatory applications.
                                       26

-------
6.   SUMMARY AND CONCLUSIONS
     Turner's method for estimating the P-G stability categories provides a
practical procedure for the routine implementation of Gaussian dispersion
models if representative cloud observations are available.  The proposed SRDT
method uses on-site meteorological data without the need for such cloud
observations, while retaining the basic structure and rationale of Turner's
method.
     A comparative analysis was performed using on-site data from three sites:
Kincaid, IL  (7 April - 31 August 1980), Longview, WA  (January - December 1991),
and Bloomington, IN  (July 1991 - July  1992).  Meteorological data included 10m
wind speed, total solar radiation, and temperature difference (AT).  All three
sites had AT data from 2-10m; 10-50m AT data were available from Kincaid and
Longview.  Valid observations from all three sites were pooled to yield a
composite data base of 19,540 hours.   The SRDT method was developed empirically
to emulate the results obtained using  the Turner stability estimation scheme.
Through iterations, optimum "cutpoints"  (meteorological parameter limits) were
determined, initially using only the 2-10m  AT data.  For the pooled data,
stability categories estimated by both methods coincided for 62 percent of the
hours, and were within one class for 89 percent.  Using the same cutpoints for
each site gave comparable results.  In an effort to evaluate the system for AT
intervals other than 2-10m, the 10-50m AT data were evaluated for two sites.
Using the same optimum cutpoints determined for the 2-10m AT data produced
virtually the same distribution of stability categories.
     Using ISC2, an analysis of the effects on design concentration ratios,
XSROT/^Tumef was completed for three hypothetical sources.  For three sites, the
ratios averaged 1.06 to 1.24 across four averaging times, three source heights,
and two concentration types.  Computed mixing heights were examined to help
understand their influence in the model-predicted concentrations.
                                       27

-------
7.   REFERENCES

Bowen, B.M., J.M. Dewart and A.I. Chen, 1983.  Stability class determination: a
      comparison for one site.  Proceedings, Sixth Symposium on Turbulence and
      Diffusion, American Meteorological Society, Boston, MA; pp. 211-214.

Bowen, B.M. and S.E. Pamp, 1994.  Comparison and Evaluation of Turbulence
      Estimation Schemes at the Rocky Flats Plant.  Proceedings, Eighth Joint
      Conference on Air Pollution Meteorology, American Meteorological Society/
      Air & Waste Management Association, 24-28 January 1994, Nashville, TN;
      pp. 54-60.

Chandler, T.J., 1965.  The Climate of London.  Hutchinson and Co., Ltd.;
      292 pp.

EPA, 1986.  Guideline on Air Quality Models  (Revised), Supplement A  (1987) and
      Supplement B  (1993) .  EPA-450/2-78-027R.  Codified as Appendix W of 40
      CFR Part 51.  U.S. Environmental Protection Agency, Research Triangle
      Park, NC.

EPA, I987a.  On-site Meteorological Program Guidance for Regulatory Modeling
      Applications.  EPA-450/4-87-013.  U.S. Environmental Protection Agency/
      Research Triangle Park, NC.

EPA, 1987b.  Ambient Monitoring Guidelines for Prevention of Significant
      Deterioration  (PSD).  EPA-450/4-87-007.  U.S. Environmental Protection
      Agency, Research Triangle Park, NC.

EPA, 1990.  Summary of Public Comments and EPA Responses on the Fourth
      Conference on Air Quality Modeling - October 1988.  EPA Docket A-88-04,
      Item No. II-O-l.  U.S. Environmental Protection Agency, Research Triangle
      Park, NC.

EPA, 1993.  Summary of Public Comments and EPA Responses on the Fifth
      Conference on Air Quality Modeling: March 1993.  EPA Docket A-88-04, Item
      No. V-C-1.  U.S. Environmental Protection Agency/ Research Triangle Park,
      NC.

EPRI, 1983.  A Catalog of Data for the Plume Model Validation and Development
      Data Base--Plains Site.  EPRI Report No. EA-3080, Electric Power Research
      Institute, Palo Alto, CA.

Gifford, F.A. Jr., 1961.  Use of routine meteorological observations for
      estimating atmospheric dispersion.  Nuclear Safety 2(4): 47-57.

Irwin, J.S., S.E. Gryning, A.A.M. Holtslag and B. Sivertsen, 1985.  Atmospheric
      Dispersion Modeling Based on Boundary Layer Parameterization.  EPA-600/
      3-85-056.  U.S. Environmental Protection Agency, Research Triangle Park,
      NC.

Irwin, J.S., J.O. Paumier and R.W. Erode, 1988.  Meteorological Processor for
      Regulatory Models (MPRM 1.2) User's Guide.  EPA-600/3-88-043R.  U.S.
      Environmental Protection Agency, Research Triangle Park, NC.

Lee, R.F., J.A. Tikvart, J.L. Dicke and R.W. Fisher, 1979.  The effect of
      revised dispersion parameters on concentration estimates.  Proceedings,
      Fourth Symposium on Turbulence, Diffusion, and Air Pollution, American
      Meteorological Society, Boston, MA; pp. 70-74.

Ludwig, F.L. and W.F. Dabberdt, 1972.  Evaluation of the APRAC-1A urban
      diffusion model for carbon monoxide.  Final report, Coordinating Research
      Council and EPA Contract CAPA-3-68 (1-69), Stanford Research Institute,
      Menlo Park, CA; 147 pp.   (NTIS No. PB 210819)
                                          28

-------
Ludwig, F.L. and W.F. Dabberdt, 1976.  Comparison of two practical stability
      classification schemes in an urban application.  Journal of Applied
      Meteorology, 15: 1172-1176.

Miller, D.H., 1981.  Energy at the Surface of the Earth. Volume 27.
      International Geophysics Series.  Academic Press; 516pp.

Pasquill, F., 1961.  The estimation of the dispersion of windborne material.
      Meteorological Magazine, 90: 33-49.

Smith, F.B., 1972.  A scheme for estimating the vertical dispersion of plumes
      from a source near ground level.  Proc. Third Meeting Expert Panel on Air
      Pollution Modeling.  NATO Committee on Challenges of Modern Society,
      XVII-1 to XVII-14.

Turner, D.B., 1964.  A diffusion model for an urban area.  Journal of Applied
      Meteorology, 3: 83-91.
                                          29

-------
          Appendix A
Results of Randomization Analysis
              A-l

-------
     As discussed in the Section 3.2,  the composite data were randomly split

into two complementary subsets and the same stability classification and

comparison applied to each.  Results of the stability calculations for the two

methods are presented in Table A-l.  The results shown are representative of

what was seen throughout all analyses performed.  Different seed values"  would

result in different cases being selected for the two bins  (i.e., Bin 0 and

Bin 1).  The results shown indicate only minor differences in the comparison

statistics between the two stability estimation methods.

     For Bin 0,  valid data for use in joint stability calculations were

available for 9834 out of 10970 hours  (89.6 percent) randomly selected of the

909-day period.  For Bin 1, valid data were available for 9706 out of 10846

hours  (89.4 percent) so selected.  For Bin 0, the stability classifications for

the two methods coincided for 62 percent of the hours and were within one

category for 89 percent of the hours.  The unstable category was the same,

while the neutral category decreased slightly and the stable category increased

slightly.  The mean absolute residual  (see Section 4.1) was 0.88% over all

categories; for daytime categories it was 0.40%, while for nighttime categories

it was 1.5%.  For Bin 1, the stability classifications for the two methods also

coincided for 62 percent of the hours and were within one category for 89

percent of the hours.  Stable and unstable categories decreased slightly, while

the neutral category increased slightly.  The mean absolute residual was 0.82%

over all categories; for daytime categories it was 0.45%, while for nighttime

categories it was 1.3%.  The frequency distributions of stability categories

for the two methods  (both bins) are displayed in Table A-l.
"Randomization was accomplished using a random number generator.  The unique
assignment of a subset to one bin versus that  assigned  to  the  other  is
controlled by the  seed value.

                                       A-2

-------
Table A-l.  Comparison of hourly stability categories via Turner versus SRDT for random  subsets of  the
            composite data  (see page A-4 for notes).a
TESTOPL*
Turner
A B C D^ D^ E F&G TOTAL
A 45 72
B 62 627
C 18 235
D^ 2 128
SRDT D 	 	 	 —
EDO
F 0 0
TOTAL 127 1062
STABILITY CLASS
A
B
C
D
E
F
UNSTABLE
NEUTRAL
STABLE
i i
202 127
489 517
488 1875
0 0
0 0
0 0
1180 2520
0
0
0
0
1079
359
446
1884
0 0
0 0
0 0
0 1
427 324
134 325
29 1822
590 2471
119
1018
1259
2493
1830
818
2297
9834
FREQUENCY (%)d
Turner SRDT
1.3
10.8
12.0
44.8
6.0
25.1
24.1
44.9
31.1
1.2
10.4
12.8
44.0
8.3
23.4
24.4
44.0
31.7
TEST1PU
Turner
A B C D.. D^, E F&G
A 63
B 56
C 13
D^ 2
SRDT
D^ 0
E 0
F 0
TOTAL 134
88
603
194
116
0
0
0
1001
STABILITY CLASS
A
B
C
D
E
F
UNSTABLE
NEUTRAL
STABLE
2 1
191 125
481 479
549 1912
0 0
0 0
0 0
1223 2517
0
0
0
0
1006
300
446
1752
0 0
0 0
0 0
0 0
478 327
116 337
26 1795
620 2459
FREQUENCY (%)e
TOTAL
154
975
1167
2579
1811
753
2267
9706

Turner SRDT
1.4
10.3
12.6
44.0
6.4
25.3
24.3
44.0
31.7
1.6
10.0
12.0
45.2
7.8
23.4
23.7
45.2
31.1
                                                     A-3

-------
                                            Notes for Table A-l

a)     On-site data are pooled from all three sites (see  Section 3.2).

b)     This analysis was done using TESTOPL,  which selects records randomly assigned an index of 0 from
      pooled data sets.  Of 21816 records read from the  3 raw meteorological input files,  10846 were ignored
      while 10970 were randomly selected for processing.  Of those selected, 1136 were rejected for missing
      data.  These included 649 with flags for invalid P-G stabilities and 1106 with flags for missing on-
      site data, including:

                                       706 with flags for 10m wind speed;
                                       682 with flags for total solar radiation; and
                                       265 with flags for 2-10m AT/AZ measurements.

  Thus, the Turner/SRDT  comparison matrix is based on 9834 valid  records  (hours),  or  89.6%  of  the  records
  randomly selected with initial  seed value: 1500.  Randomly selected were  5679  daytime hours  and  5291
  nighttime hours; processed were 4889 daytime hours and 4945 nighttime hours.

c)     This analysis was done using TEST1PL,  which selects records randomly assigned an index of 1 from
      pooled data sets.  Of 21816 records read from the 3 raw meteorological input files,  10970 were ignored
      while 10846 were randomly selected for processing.  Of those selected, 1140 were rejected for missing
      data.  These included 646 with flags for invalid P-G stabilities and 1122 with flags for missing on-
      site data, including:

                                       715 with flags for 10m wind speed;
                                       699 with flags for total solar radiation; and
                                       264 with flags for 2-10m AT/AZ measurements.

  Thus, the Turner/SRDT  comparison matrix is based on 9706 valid  records  (hours),  or  89.5%  of  the  records
  randomly selected with initial  seed value: 1500.  Randomly selected were  5686  daytime hours  and  5160
  nighttime hours; processed were 4875 daytime hours and 4831 nighttime hours.

d)     Using TESTOPL,  the stability classifications for the two methods coincided for 61.7% of  the hours, and
      were within one category for 89.4% of the hours (with P-G categories F and G via Turner  combined).

e)     Using TEST1PL,  the stability classifications for the two methods coincided for 61.6% of  the hours, and
      were within one category for 89.4% of the hours (with P-G categories F and G via Turner  combined).
                                                     A-4

-------
             Appendix B
Results of Gaussian Dispersion Modeling:



        A Consequence Analysis
                  B-l

-------
Table B-l.  Design concentration ratios  derived  from  ISC2ST  for  Bloomington,
            IN site; 2-10m AT  (see  Section  5.1).
Sourceb
Single
35m
Stack
Single
100m
Stack
Single
200m
Stack
Ambient Concentration" via ISC2ST (/*gnf3)
Avg. Time Turner0 SRDTd /„ Rat/*°" ,
3 ^XsRHT/ATunier'
1-hour
3 -hour
24 -hour
Period
(8437 hours)
1-hour
3 -hour
24 -hour
Period
(8437 hours)
1-hour
3 -hour
24 -hour
Period
(8437 hours)
245.7
232.9
217.3
198.4
76.1
67.9
4.38
4.12
55.9
55.3
34.2
31.1
7.98
6.49
0.349
0.348
30.2
24.3
15.4
12.0
2.93
2.60
0.104
0.097
238.1
233.7
225.3
214.2
75.7
67.9
4.90
4.64
56.8
47.3
35.5
33.7
9.26
8.02
0.391
0.376
30.7
25.9
17.7
11.2
3.24
2.65
0.110
0.105
0.97
1.00
1.04
1.08
1.00
1.00
1.12
1.13
1.02
0.85
1.04
1.08
1.16
1.24
1.12
1.08
1.02
1.06
1.15
0.94
1.10
1.02
1.06
1.08
"For  each  averaging  time,  high 1st  high (HlH)  concentration appears above
dotted line; high 2nd high  (H2H) concentration appears below dotted line.

bSee  text,  Section 5.0,  for description of all source and receptor parameters
used in the dispersion model runs.

"Hourly mixing  heights are computed based on Turner-derived stabilities.

dHourly mixing  heights are computed based on SRDT-derived stabilities.
'Statistical  analysis  (see  Section 5.1):   median
                 s
 For all 24 values:
 For 12 HlH concentration ratios only:
 For 12 H2H concentration ratios only:
1.06      1.06   0.08  1.05  1.08
1.05      1.07   0.06  1.07  1.06
1.07      1.05   0.10  1.04  1.10
                                      B-2

-------
Table B-2.   Design concentration ratios derived from ISC2ST for Kincaid, IL
             site;  2-10m AT (see Section 5.2).
Source11
Single
35m
Stack
Single
100m
Stack
Single
200m
Stack
Ambient Concentration" via ISC2ST (pgm3)
Avq. Time Turner0 SRDTd /„ Ra5i°e \
^XSRDT/ ATun*r>
1-hour
3 -hour
24 -hour
Period
(2842 hours)
1-hour
3 -hour
24 -hour
Period
(2842 hours)
1-hour
3 - hour
24 -hour
Period
(2842 hours)
254.4
233.5
201.9
186.4
60. 8f
53.5
5.86
5.25
61.8
54.7
37.2
27.8
9.21
7.43
0.649
0.649
26.3
24.6
12.3
11.1
3.00
2.44
0.225
0.221
236.7
236.1
234.5
194.4
69.8
56.5
6.43
5.96
61.8
37.0
36.6
30.6
10.6
7.48
0.567
0.563
32.2
30.1
19.9
15.7
3.65
2.70
0.172
0.166
0.93
1.01
1.16
1.04
1.15
1.06
1.10
1.13
1.00
0.68
0.99
1.10
1.15
1.01
0.87
0.87
1.22
1.22
1.62
1.41
1.22
1.11
0.77
0.75
"For  each averaging time,  high 1st high (H1H)  concentration appears above
dotted line; high 2nd high  (H2H)  concentration appears below dotted  line.

bSee  text,  Section 5.0,  for description of all source and receptor parameters
used in the dispersion model runs.

"Hourly mixing  heights are computed based on Turner-derived stabilities.

dHourly mixing  heights are computed based on SRDT-derived stabilities.
"Statistical  analysis  (see  Section 5.2):    median
                 s
 For all 24 values:
 For 12 H1H concentration ratios only:
 For 12 H2H concentration ratios only:

fOne hour was missing  in the  computation  of  this  concentration.
1.08      1.07   0.21  1.05  1.22
1.12      1.10   0.22  1.08  1.21
1.05      1.03   0.20  1.01  1.22
                                      B-3

-------
 Table B-3.   Design concentration ratios derived from ISC2ST for Kincaid, IL
             site; 10-50m AT  (see Section 5.2).
Sourceb
Single
35m
Stack
Single
100m
Stack
Single
200m
Stack
Ambient Concentration" via ISC2ST (ngm*)
Avg. Time Turner0 SRDT" ,v Ra5i°C ^
>XSRDT/ ATunBr'
1-hour
3 - hour
24 -hour
Period
(2842 hours)
1 - hour
3 - hour
24 -hour
Period
(2842 hours)
1-hour
3 -hour
24 -hour
Period
(2842 hours)
254.4
233.5
201.9
186.4
60. 8f
53.5
5.86
5.25
61.8
54.7
37.2
27.8
9.21
7.43
0.649
0.649
26.3
24.6
12.3
11.1
3.00
2.44
0.225
0.221
236.7
236.1
234.5
194.4
69.8
56.5
6.43
5.96
61.8
37.0
36.6
30.6
10.6
7.48
0.567
0.563
32.2
30.1
19.9
15.7
3.65
2.70
0.172
0.166
0.93
1.01
1.16
1.04
1.15
1.06
1.10
1.14
1.00
0.68
0.99
1.10
1.15
1.01
0.87
0.87
1.22
1.22
1.62
1.41
1.22
1.11
0.77
0.75
"For each averaging time,  high 1st high (H1H)  concentration appears above
dotted line; high 2nd high  (H2H)  concentration appears below dotted  line.

bSee text,  Section 5.0,  for description of all source and receptor parameters
used in the dispersion model runs.

"Hourly mixing  heights are computed based on Turner-derived stabilities.

dHourly mixing  heights are computed based on SRDT-derived stabilities.
"Statistical  analysis  (see  Section 5.2):    median
                                                                         B
                                            1.08
                                            1.12
                                            1.05
 For all 24 values:
 For 12 H1H concentration ratios only:
 For 12 H2H concentration ratios only:

fOne hour was missing in the computation  of  this  concentration.

                                      B-4
1.07  0.21  1.05   1.22
1.10  0.22  1.08   1.21
1.03  0.20  1.01   1.22

-------
Table B-4.   Design concentration ratios derived from ISC2ST for Longview, WA
             site;  2-10m AT (see Section 5.3).

Source1"



Single
^Sm
jjiii
Stack





GJ»-.^1x-v
oingie
100m
1 V/Vslll
Stack






Single
200m
^\j\jiii
Stack



Ambi
Avg . Time



3 - hour


Period
(8187 hours)






Period
(8187 hours)






Period
(8187 hours)
ant Concentratia
Turner"
234.4
228.8
196.8
189.6
82.7
58.5
7.24
6.14
55.9
55.8
34.3
29.6
7.01
6.11
0.613
0.575
30.2
26.3
17.8
10.6
2.56
2.39
0.195
0.187
n" via ISC2ST (/i
SRDTd
236.3
235.5
217.7
211.5
82.7
58.5
9.31
7.34
55.9
55.8
34.3
33.1
8.58
7.76
0.962
0.835
36.0
32.5
23.7
17.1
4.34
2.88
0.306
0.301
gnf5)
Ratio"
(XSRDT/XTUTOI-)
1.01
1.03
1.11
1.12
1.00
1.00
1.29
1.18
1.00
1.00
1.00
1.12
1.22
1.27
1.57
1.45
1.19
1.24
1.33
1.62
1.70
1.21
1.57
1.61
"For  each  averaging time,  high 1st high (H1H)  concentration appears above
dotted line; high 2nd high  (H2H)  concentration appears below dotted  line.

bSee  text,  Section 5.0,  for description of all source and receptor parameters
used in the dispersion model runs.

"Hourly mixing  heights are computed based on Turner-derived stabilities.

dHourly mixing  heights are computed based on SRDT-derived stabilities.

"Statistical analysis  (see Section 5.3):    median     X
                                                             s
 For all 24 values:
 For 12 H1H concentration ratios only:
 For 12 H2H concentration ratios only:
1.20      1.24   0.23  1.22  1.19
1.20      1.25   0.25  1.23  1.21
1.19      1.24   0.22  1.22  1.18
                                      B-5

-------
Table B-5.   Design concentration ratios derived from ISC2ST for Longview, WA
             site;  10-50m AT (see Section 5.3).
Source1"
Single
35m
Stack
Single
100m
Stack
Single
200m
Stack
Ambient Concentration" via ISC2ST (ngm3)
Avg. Time Turner* SKDT* (XWxL)
1-hour
3 -hour
24 -hour
Period
(8187 hours)
1-hour
3 -hour
24 -hour
Period
(8187 hours)
1-hour
3 -hour
24 -hour
Period
(8187 hours)
234.4
228.8
196.8
189.6
82.7
58.5
7.24
6.14
55.9
55.8
34.3
29.6
7.01
6.11
0.613
0.575
30.2
26.3
17.8
10.6
2.56
2.39
0.195
0.187
236.3
235.5
217.7
211.5
82.7
58.5
9.32
7.35
55.9
55.8
34.3
33.1
8.58
7.78
0.963
0.835
36.0
32.5
23.7
17.1
4.34
2.88
0.306
0.301
1.01
1.03
1.11
1.12
1.00
1.00
1.29
1.20
1.00
1.00
1.00
1.12
1.22
1.27
1.57
1.45
1.19
1.24
1.33
1.62
1.70
1.21
1.57
1.61
"For  each averaging time,  high 1st high (H1H)  concentration appears above
dotted line; high 2nd high  (H2H)  concentration appears below dotted  line.

bSee  text,  Section 5.0,  for description of all source and receptor parameters
used in the dispersion model runs.

"Hourly mixing  heights  are computed based on Turner-derived stabilities.

dHourly mixing  heights  are computed based on SRDT-derived stabilities.
'Statistical  analysis  (see  Section 5.3):    median
                 s
 For all 24 values:
 For 12 H1H concentration ratios only:
 For 12 H2H concentration ratios only:
                             sc
1.20      1.24   0.23  1.23  1.19
1.20      1.25   0.25  1.23  1.21
1.19      1.24   0.22  1.22  1.18
                                      B-6

-------
   3, ODD
   2, 5DO
   2, ODD
    , 500
    J ODD
     500
               7    B
                             10   11    12    13   14   15   16   17    18   19   20

                              HOUR  CDay-time  on|
  Figure B-la.  Mean mixing  height by hour of the  day for Bloomington,  IN site;
                2-10m  AT  (see  Table  B-l).
N
   3, ODD
  2,50O -
  2,000  —
   1J5OO  —
     000  —
     5OO  —
                    8    9    10   11    12   13   14   15   16   17   18   19   20
                              HOUR  CDaytime  only^

  Figure B-lb.  Median mixing height  by hour of the  day for Bloomington,  IN
                site; 2-10m AT  (see Table B-l).
                                        B-7

-------
N
   2_ 5DO
   2^-400
   2, 2OO
   2, ODD
    , BOO
    J BDO
    , 4DO
     2DO
     ODD
          5    6    7    8    9   ID   11   12   13    14   15   16   17   18   19
                               HOUR  C Day-time  only]}

  Figure  B-2a.   Mean mixing height by hour of the day for Kincaid, XL  site;
                 2-10m AT (see Table B-2).
Kl
   3, ODD
   2, 500
    ^ ODD
    J 5OO
     000
     500
-O
                             J_
                                            J
                                                 L
          5    B    7    8    9    1O   11   12   13    14   15   16   17   18   19
                              HOUR  CDaytime only}

  Figure B-2b.  Median mixing height by hour of the day for Kincaid, IL site;
                2-10m AT (see Table B-2).
                                        B-8

-------
N
   2, EDO
   2, 4OO
   2, 2OD
   2, ODD
    ,800  —
    , BOO  -
   1,400  —
   1,200  —
     000
                              _L
                                   _L
                                        _L
                                             _L
                                                  J_
                                                       _L
                                                           J_
                                                                _L
                                                                          _L
           5     6    7    B    9    1O   11    12   13   14   15   16   17   18   19
                               HOUR  CDaytime  only}

   Figure B-3a.  Mean mixing height  by hour of  the  day for Kincaid, IL  site;
                 10-50m AT  (see Table  B-3).
         -e
M
   3, OOO
2, 500
   2, OOO
    , 500
    , OOO
     500 —
                              _L
                                            J_
                                                           J_
                                                                _L
                                                                          _L
                           9    1O   11    12    13   14   15   16   17   18
                            HOUR  CDaytime onlyD
                                                                               19
  Figure B-3b.   Median mixing height by hour of  the  day for Kincaid, IL site;
                 10-50m AT (see Table B-3).
                                         B-9

-------
N
   3J ODD
   2. 500
   2, ODD
    , 5OO
   -1 „ ODD
     500
          J	I
                                      SRDT    Turner
                                   \	|	I	I
          5    B    7    B    9   1O   11   12   13    14   15   16   17   18   19

                               HOUR  CDaytime on Iy^


   Figure B-4a.   Mean mixing height by hour of the day for Longview, WA site;
                 2-10m AT (see Table B-4).
   3, OOO
   2,5OO -
   2,OOO —
_y 1_ 5OO

M
   1,000 —
     5OO — -
                                      SRDT    Turner-
                                       e
          5    B    7    8    9    1O   11   12    13
                                                          15   16   17   18   19
                              HOUR  CDaytime only^
  Figure B-4b.  Median mixing height by hour of the day for Longview, WA site;
                2-10m AT (see Table B-4).
                                        B-10

-------
N
  2, 6DO
    ^ -4DO
  2, 2DD
  2, OOO
    , BDD
    _ BOO
     4DO
   1 _ 2OO
          J	L
                                  J	L
                                                 J	L
          5    6    7    8    9   1O   11   12   13    14   15   16   17   18   19
                               HOUR  CDaytime on Iy}

  Figure B-5a.  Mean mixing height by hour of the day for Longview, WA  site;
                10-50m AT (see Table B-5).
N
   3, OOO
   2, 500
   2, OOO
    j 5OO
    , 000
     500
          J	I
                                  J	I
          5    6    7    8    9   1O   11   12   13    1-4   15   16   17   18   19

                               HOUR  CDayt_ime only^i

  Figure B-5b.  Median mixing height by hour of the day for Longview, WA  site;
                10-50m AT (see Table B-5).
                                        B-ll

-------
                                     TECHNICAL REPORT DATA
                                (Please read Instructions on reverse before completing)
  1. REPORT NO.
    EPA-454/R-93-055
                  3. RECIPIENT'S ACCESSION NO.
  4. TITLE AND SUBTITLE
    Evaluation of a Solar Radiation/Delta-T Method for Estimating
    Pasquill-Gifford (P-G) Stability Categories
                                                                    5. REPORT DATE
                                                                      October 1993
                  6. PERFORMING ORGANIZATION CODE
  7. AUTHOR(S)
    C. Thomas Coulter
                  8. PERFORMING ORGANIZATION REPORT NO.
  9. PERFORMING ORGANIZATION NAME AND ADDRESS

    U.S. Environmental Protection Agency
    Office of Air Quality Planning and Standards
    Technical Support Division
    Research Triangle Park, NC  27711
                  10. PROGRAM ELEMENT NO.
                  11. CONTRACT/GRANT NO.
  12. SPONSORING AGENCY NAME AND ADDRESS
                                                                    13. TYPE OF REPORT AND PERIOD COVERED
                                                                    14. SPONSORING AGENCY CODE
  15. SUPPLEMENTARY NOTES
  16. ABSTRACT
      This publication documents the effort made to develop and evaluate a new methodology for
  estimating stability category using on-site meteorological data that can be automatically collected and
  logged,  e.g., wind speed and solar radiation during daytime and temperature difference (AT) at night.
  The new method (Solar Radiation/Delta-T, SRDT) uses 5 wind speed classes and 4 insolation classes
  during daytime, and 3 wind speed classes and 2 AT classes during nighttime.  To fulfill the objectives of
  the evaluation three on-site meteorological data bases were obtained: Kincaid, IL (4/80 - 8/80),
  Longview, WA (1/91 - 12/91), and Bloomington, IN (7/91 - 7/92).  The data were pooled to yield
  19,540 valid hours.  Using the composite data, stability classification criteria were determined iteratively
  for the SRDT method.  Stability categories via both methods were rigorously compared for all valid
  hours.  Overall, stability categories coincided for 62 % of the hours examined, and were ± 1 class for
  89 % of the hours.  The same criteria were then applied to each of the three sites individually  to assess
  site to site variability.  This variability was seen to be of the order of that seen within an individual site.
  For two sites, the SRDT method was evaluated for AT values measured from an interval other than 2-
  10m (i.e., 10-50m).  The same methodology produced vurtually identical results.  Finally, a dispersion
  model (ISC2) was run to demonstrate the effect of the SRDT method on design  concentration  ratios
  using meteorological data from all sites and for both AT measurement heights.
  17.
                                      KEY WORDS AND DOCUMENT ANALYSIS
                    DESCRIPTORS
                                                  b. IDENTIFIERS/OPEN ENDED TERMS
                                                                                      c. COSATI Field/Group
    Air Pollution
    Atmospheric Dispersion Modeling
    Meteorological Preprocessors
    Atmospheric Stability Estimation
 18. DISTRIBUTION STATEMENT


    Release Unlimited
19. SECURITY CLASS (Report)
   Unclassified
21. NO. OF PAGES
       40
                                                  20. SECURITY CLASS (Page)
                                                     Unclassified
                                    22. PRICE
EPA Form 2220-1 (Rev. 4-77)   PREVIOUS EDITION IS OBSOLETE
    U.S. Environme;V.::   ; '•cclion Agency
    Region 5, Library •;; L-I2J)
    77 West Jackson Boulevard,  12th Floor
    Chicago, IL  bOc.?'4"3,rOO

-------