-------
                                  DISCLAIMER

     The Information 1n this document  has  been  funded wholly or in part by
the U.S. Environmental Protection Agency under  contract 808485-02 to the
Geophysical Institute, University of Alaska,  Fairbanks.  It has been subjected
to the Agency's peer and administrative  review,  and 1t has been approved for
distribution.  Mention of trade names  or commercial products does not constitute
endorsement or recommendation for use.
                                   \
                                      11

-------
                                   ABSTRACT

     High winter carbon monoxide levels 1n Anchorage,  as  in  Fairbanks,  are
due to Intense nocturnal (ground-based) inversions  persisting  through the
periods of maximum emissions and at times  throughout the  day.   The  problem
1s exacerbated by the large amounts of carbon monoxide emitted during cold
starts at low temperatures.  The Anchorage situation is unusual  in  that the
nocturnal inversion develops most often with  a substantial north-south
pressure gradient and easterly geostrophic winds.   The Chugach Range to the
east sometimes produces a "wind shadow" effect in the  city,  and  almost  all
the CO violations examined occured in these conditions.  There is evidence
that inversions are significantly stronger, and dispersion conditions probably
worse, near the mountain front than at the airport  weather observation station,
CO forecasting in Anchorage would require  close cooperation  between the U.S.
NOAA Weather Service and the Municipality; improvement in communications
between the Fairbanks North Star Borough and  the Weather  Service is also
essential 1f the quality of the Fairbanks  CO  forecasts is to be  improved.
Measurements of mixing heights 1n Fairbanks suggest that  a mixing height of
10 m be considered the maximum for worst-case modeling of surface-source
pollutants; values as low as 6 m were observed. As an interim measure,
similar values are recommended for Anchorage.

     This report was submitted in fulfullment of grant 808485-02 by the
Geophysical Institute, University of Alaska,  Fairbanks under sponsorship
of the U.S. Environmental Protection Agency.   This  report covers analysis of
existing data and new data collected in late 1981,  1982,  and early  1983.  The
study was completed September 1983.
                                     111

-------
                              CONTENTS

Abstract	      iii
Figures	        v
Tables	,	     viii
     1.   Executive Summary 	        1
     2.   The Problem	        3
               Discovery of the CO problem	        3
               Fairbanks	        3
               Anchorage	        5
               Research Plan	        7
     3.   Fairbanks CO Forecasts	        9
               Forecasting method 	        9
               Forecast performance 	       10
               Recommendations	       20
     4.   Anchorage	       21
               Introduction 	       21
               Observed carbon monoxide levels	       21
               Meteorology	       25
               Traffic and seasonality	       29
               Vertical temperature structure 	       31
               The wind field	       36
               The larger basin:  inner Cook Inle»t	       37
               Conclusions	       36
     5.   Mixing Heights and Modelling	       40
               Introduction	       40
               Effects of "tuning" models 	       40
               Measurement of Fairbanks mixing heights.  .  .       41
               Anchorage mixing heights 	       48
               Conclusions	       48
     6.   Summary and Conclusions	       49
               Conclusions	       49
               Recommendations	       50

References	       52

Appendix 1:  Modifications necessary to use standard dispersion    54
             models at high latitudes 	
Appendix 2:  The influence of the form of the temperature
             sounding on the depth of the mixing layer produced
             over a city	       67
                                      iv

-------
                              CONTENTS

Abstract	       -Hi
Figures . ,	         v
Tables	„	      viii
     1.   Executive Summary 	         1
     2.   The Problem	         3
               Discovery of the CO problem.	         3
               Fairbanks	         3
               Anchorage	         5
               Research Plan	         7
     3.   Fairbanks CO Forecasts	         9
               Forecasting method 	         9
               Forecast performance 	        10
               Recommendations	        20
     4.   Anchorage	        21
               Introduction 	        21
               Observed carbon monoxide levels	        21
               Meteorology	        25
               Traffic and seasonality	        29
               Vertical temperature structure 	        31
               The wind field	        36
               The larger basin:  inner Cook Inleit	        37
               Conclusions	        38
     5.   Mixing Heights and Modelling	        40
               Introduction 	        40
               Effects of "tuning" models 	        40
               Measurement of Fairbanks mixing heights.  .  .        41
               Anchorage mixing heights	        48
               Conclusions	        48
     6.   Summary and Conclusions	        49
               Conclusions	        49
               Recommendations	        50

References	        52

Appendix 1:  Modifications necessary to use  standard  dispersion    54
             models at high latitudes 	
Appendix 2:  The influence of the form of the temperature
             sounding on the depth of the mixing  layer produced
             over a city	        67
                                      1v

-------
                                   FIGURES


Number

2-1   Landsat photo of Fairbanks area	    4

2-2   Map of major Anchorage streets, showing CO monitoring stations.
      7C = 7th and C, B5 = Benson and Spenard, GS = Garden  Site,
      and SL = Sand Lake. ,B = Bus Barn, site of Tethersonde ascent,  .    6

3-1   7:00 am forecast of day's highest 8-hour CO level  against maxi-
      mum 8-hour level observed the same day, 1979-1980.  Numbered
      points are in Table 1	11

3-2   6:30 am forecast of day's highest 8-hour CO level  against maxi-
      mum 8-hour level observed the same day, 1980-1981.  Numbered
      points are in Table 1	12

3-3   6:00 am forecast of day's highest 8-hour CO level  against maxi-
      mum 8-hour level observed the same day, 1982-1983.  Numbered
      points are in Table 1; I's are ice fog days	   13

3-4   6:00 am persistance forecast of day's  highest 8-hour  CO  level
      against the highest 8-hour CO level  actually  observed  the same
      day, 1982-1983.  Numbered and lettered points match those in
      Figure 3-3	16

3-5   Morning dispersion forecast against  maximum 8-hour level of
      CO observed the same day, 1982-1983.  Dispersion plotted is
      the lowest forecast for the day;  i.e., a forecast  of  good
      becoming poor is plotted as poor.  I's are ice foy cases ....   17

3-6   Individual plots of cases with observed 8-hour CO  levels 15 ppm
      or more forecast at less than 9 ppm.  Solid line 1-hour  average
      CO, dashed line temperature.  Cloud  cover is  in tenths,  ceiling
      in thousands of feet,  and windspeed  in knots, following  NOAA
      Weather Service usage	18

4-1   Composite Landsat photo of the Cook  Inlet area	22

4-2   NOAA-4 thermal Infrared Image of  the Cook Inlet area.  Anchorage
      is at ANC, and light tones are cold	23

4-3   Lower curves:  7th & C (solid) Benson and Spenard   (dashed)
      and Garden Site (dotted) hourly CO values.  Middle curve:
      100-m temperature minus surface temperature (100-m Inversion
      strength) at Anchorage airport.  Upper curves:  Airport
      (dot-dashed) Benson and Spenard (dashed) and Garden S1tt»
      (dotted) surface temperatures	24

-------
                                   FIGURES

Number                                                                 Page


4-4   Traffic counts (averaged over 5-8 days)  fo&«Benson  and  Spenard
      (solid line) and Sand Lake (dashed line).   Based  on December
      1982 and January 1983 data	T	    24

4-5   Lower curves:  Normalized hourly CO values  for  same stations
      and time periods as Figure 4-3.  Middle  curve:  cloud cover;
      solid area cloud, open area fog.  Top curve:  windspeed.  ....    25

4-6   NOAA-6  thermal infrared image,   16 Feb 1980.  8:39 am  ....    26

4-7   TIROS-N thermal infrared image,  18 Feb  1980.   1:35 pm	    28

4-8   Weather map, 2 pm AST (Alaska Standard Time)  3  Dec. 1982.  ...    29

4-9   Weather map, 2 pm AST 16 Feb. 1980	    29

4-10   February mean hourly values of CO for the  three  stations  with
       the longest periods of punched data	    30

4-11   December mean hourly values of CO for the  same three stations
       as Figure 4-10	    32

4-12   Airport soundings for a high CO episode in December 1982.
       Heavy lines are 1 pm soundings.  The first number  after
       each day and time is the corresponding  hourly  mean CO  (ppm)
       at Benson and Spenard; the second is the surface temperature
       at Benson and Spenaro.	    33

4-13   Possible soundings at Benson and Spenard for 1 pm  Dec.  25,
       1982	    34

4-14   Possible paths for a1 r flow over the Chugach Range	    35

4-15   Comparison of an Inland Tethersonde sounding made  at the  Bus
       Barn (B on Figure 2-2) with flanking soundings from the air-
       port 	    36

4-16   Comparison of wind directions measured  simultaneously  at  the
       airport (solid line), 7th and C (dashed line)  and  Tudor and
       Lake Otis (dotted line)	    36

5-1    Main road net 1n Fairbanks.  CF (Creamer's Field)  is site
       of background ascent;  D (downtown) 1s  the dty  ascent
       location	    4?
                                 v1

-------
                              FIGURES
Number
5-2	.£0 levels near the times of the three  sets  of ascents. . . T.   44
5-3    Background and downtown soundings,  15  December 1981	   45
5-4    Background and downtown soundings,  22  December
       1981	   46
5-5    Background and downtown soundings,  23  December 1981. .....   47

-------
                               TABLES
Number                                                                 Page
3-1  Poorly forecast cases examined in detail	      H
                                 V111

-------
                                 SECTION  1

                             EXECUTIVE  SUMMARY


     Urban winter air pollution problems  in Alaska  appear to be out of all
proportion to the size of the cities  involved.   Both Ancnorage  (population
around 175,000) and Fairbanks (population around 40,000) exceeded the 9 ppm
8-hour CO level 35 days or more during  the period November 1982 through
February 1983, and levels of 15 ppm were  reached twice in Anchorage and
seven times in Fairbanks.  Longer term  records  are  similar, but with less
contrast between the two cities in the  number of days with very high carbon
monoxide levels.  This study was carried  out  to clarify the meteorological
conditions responsible for these high levels  in Anchorage, to provide real
data for mixing heights in Fairbanks  and  to  evaluate the CO forecasting
scheme used in Fairbanks.

     The CO problem in Anchorage, like  that  in  Fairbanks, is due to a
combination of strong  ground inversions and  low wind  speeds which persist
through the hours of maximum emissions.  At  lower  latitudes, these
"nocturnal inversion"  conditions are confined to the  night, when emissions
are  low.  At  latitudes north of 60°N, however,  they persist throughout the
day.   In most  places,  including Fairbanks, nocturnal  inversions are
associated with  anticyclones.  In Anchorage,  however, they are  associated
with winds from the east being blocked  by the Chugach Range east of
Anchorage.  There is good reason to think that  the  eastern and  most
sheltered part of the  city is more vulnerable to pollution than is the
western part where the majority of monitoring sites are  located.

     Conventional air  pollution mouels  do not perform well in high-latitude
winter situations.  One major problem is  that the  combination of extremely
poor vertical mixing with highly variable wind  directions is normal at high
latitudes; another is  that time-averaged  wind directions frequently vary
by as much as  180° across a city, or over a  height  difference of a few
meters.  Existing models could easily be  modified  to  handle the first
problem, and this is essential if computer models  are to be used to evaluate
the  impact of changed  source distributions.   The correct handling of the
complex wind fields 1n sheltered high-latitude  cities may require designing
models for Individual  locations after collecting the  necessary  data on
the wind fields.

     Actual measurement of mixing heights within a  city  block of the CO
monitoring site 1n Fairbanks has confirmed that complete mixing is confined
to a layer which may be as little as  6  meters deep.  Some additional dis-
persion may occur from updrafts along heated  buildings,  but the deepest
mixing layer which should ever be used  for worst-case modeling  of dispersion
from surface sources 1n Fairbanks is  10 m.  (Some  past  attempts have been
based on a 100-m or even 200-m mixing layer.)  Similar  measurements  are

-------
needed 1n Anchorage, but for the present, 10 . m should be considered  a
maximum worst-case mixing heioht for that city also.   For elevated  sources,
worst-case modeling should be based on a mixing layer just deep  enough  to
include the source.

     The Fairbanks forecasting scheme has not been successful  in forecasting
CO levels above 15 ppm, but dees show a correlation coefficient  of the  order
of 0.6 between forecast and observed levels.  The single most  important
factor in Improving forecasts is better communications between the CO fore-
caster (at the Fairbanks North Star Borough) and the  NOAA Weather Service
office in Fairbanks, which provides the dispersion forecasts on  which the
CO forecasts are in part based.  Several meteorological  situations were
found to be associated consistently with high CO levels: moderate winds at
the airport while winds downtown were calm, low-level transport  of warm air
into the area, and sunset times shortly before the evening traffic  rush.

     This study was based on existing data and on that collected by  state
and local agencies.  We found that additional data were needed on the v»ind
field  and variation of  inversion strength across Anchorage. A data
collection  and interpretation program involving roughly 10 meteorological
towers  in conjunction with tethered balloon measurements is strongly
recommended for the Anchorage area.  Successful modeling of the  effects
of  changing source  distributions in the Anchorage area cannot  be expected
without  these data.

-------
                                SECTION 2

                               THE PROBLEM
DISCOVERY OF THE CO PROBLEM
     Air pollution 1n the sparsely settled  Arctic,  unlikely  as  it may at
first sight seem, 1s a very real  and serious  problem.   In Fairbanks, ice fog
was recognized by 1949 as being due to H20  released by  human activity (Oliver
and Oliver, 1949), and the continued study  of 1ce fog led to identification of
other potential pollutants (e.g.  Benson,  1965).  The presence of high lead and
halogens was confirmed shortly thereafter (Winchester et al., 1967).  The
first measurements of particulates and gaseous pollutants, however, were unre-
lated to the 1ce fog studies (Holty, 1983,  personal  communication).  Back-
ground measurements of particulates and sulfur oxides were initiated by the
Public Health Service 1n 1967 and carbon  monoxide measurements  were added in
1969.  Levels of CO were found from the start to be shockingly  high - hourly
averages of almost 70 ppm were recorded in  early years  (Holty,  1973).  CO
monitoring was taken over by the state and  the Fairbanks North  Star Borough
in the early '70's and continued  to the present.

     Once the existence of a CO problem was discovered, the  previous work on
1ce fog Identified the causes.  Fairbanks in  winter has about as close to zero
dispersion as is found in nature.  Both exceptionally strong ground-based
Inversions and very low wind speeds (Benson,  1965)  contribute to the problem.
The only larger city 1n the State, Anchorage, is notorious for  its winter
winds and, being considerably warmer, lacked  the tell-tale ice  fog.  Still, in
the early '70's a CO monitor was  installed  in a downtown location, at 7th and
C streets.  Once initial problems in calibration were overcome, it was clear
that Anchorage also violated Federal Standards for  CO.  Both Fairbanks and
Anchorage were designated non-attainment  areas in 1978.


FAIRBANKS

     The meteorology of a1r pollution 1n  Fairbanks  has  been  studied extensively
with reference to 1ce fog (Benson, 1965,  1970; Weller,  1969; Bowling 1967,
1970; Bowling et al., 1968; Fahl, 1969; Holmgren et al., 1975;  Holty, 1973;
Ohtake, 1970; Bowling and Benson, 1978; Jayaweera et al., 1975; Wendler, 1975)
and 1s reasonably well understood.  The city  1s located 1n a southward-opening
arc of hills 1n the northwest corner of a larger southwest-facing arc on the
north side of the Tanana River valley (Figure 2-1). The shelter of the hills
keeps wind speeds down, and the very low  elevation  of the winter sun  (less
than 2° above the horizon &t noon near the  winter solstice)  allows some of the
most Intense radiative Inversions 1n the  world to form  and to maintain them-
selves for days at a time.  (M1dlat1tude  cities may develop  strong ground-based
Inversions on clear, calm nights, but solar heating on  clear days will assure

-------
                 Figure 2-1.  Landsat photo of Fairbanks area
that the Inversions are at least weakened during the day when pollutant inputs
are strongest.) The problem 1s made worse by the fact that cold starts  of
automobiles produce CO levels an order of magnitude greater than the same  car
produces at warm Idle.  Given the usually short distance that a car Is  driven
on a single trip 1n a small town such as Fairbanks, these cold starts may
account for as much as  75% of the CO produced.  Furthermore, catalytic
converters do not operate efficiently until  they warm up (Leonard,  1975).
Luckily this problem 1s to some extent self-limiting, as few automobiles will
start at temperatures of -30°C or lower without preheating the engine (usually
with an electrical  heater which may act on the oil, the coolant, or the engine
block Itself).

-------
Such preheated, properly used, virtually eliminate the surge of CO of a true
cold start.

     Because of the extremely limiting meteorology, the CO problem can only  be
ameliorated by controlling emissions - primarily automotive emissions.  The
first line of attack was improvement of traffic flow in Fairbanks by the crea-
tion of a grid of one-way streets 1n the downtown area (Leonard, 1977) and,
later, by the building of the Steese Expressway Bypass.  Additional  strategies
were aimed at minimizing cold starts by encouraging the use of engine heaters
at temperatures considerably above those at which plug-ins are necessary if
the engine 1s to start, and discouraging vehicle traffic during periods  of
poor dispersion.  The  latter approach has involved development of a public
transit system, forecasting of high CO levels, and elimination of fares  on
the transit system when forecast or observed levels of CO exceed 15 ppm on an
8-hour basis.

     The  forecasting scheme used is based on the assumption that the ratio of
daily 8-hour maximum CO levels to the 8-hour average immediately proceeding
the time  the forecast  1s made will be the same as that of the previous day.
This preliminary forecast is then modified in accordance with dispersion fore-
casts made by the local office of the National Weather Service.  One of  the
aim:; of the present study was to determine the meteorological conditions which
lead to "good"  and  "bad" forecasts.  The results of this part of the project
are given in Chapter 3.

     A second objective for Fairbanks was to quantify the mixing heights occur-
ring within the city.  Previous work had established that lapse rates outside
the built-up area were normally dominated by ground-based inversions (Billelo,
1966; Benson, 1965) frequently possessing a stepped structure when examined
in detail  (Holmgren et al., 1975).  In addition, studies of the Fairbanks
heat island had established that an inversion existed at some point between
the surface and 90 m under conditions of strong background inversion (Bowling
and Benson, 1978).  Although the mixing height was estimated from the heat
island intensity to approximate 60 m, any value from 0 to almost 90 m was
consistent with the observations.  Mixing height is a critical parameter in
dispersion models, and there has been a lamentable tendency to run "worst-
case" models with a 100-m mixing height.  Therefore we used a Tethersonde
(T.M.) system to measure the lapse rate within and outside the city of
Fairbanks and compared the two to estimate mixing height.  The results are
given in  Chapter 5.


ANCHORAGE

     In contrast to Fairbanks, Anchorage was virtually unstudied before this
research  began.  As mentioned above, observations began in the early 70's at
7th and C, a block south and a couple of blocks east of the core business area

-------
              ANCHORAGE
Figure 2-2.  Map  of major Anchorage streets, showing CO monitoring stations.
7C « 7th and C, BS « Benson and Spenard, GS = Garden Site, and SL = Sand Lake,
"B * Bus Barn,  site of Tethersonde ascent.

-------
(Figure 2-2).  Although CO levels were high enough  to result  in  Anchorage
being designated a non-attainment area in 1978,  substantial  improvement  was
obtained by a change in traffic flow patterns (La More,  1983,  personal communi-
cation).  In December 1978 a second station was  added at Benson  and  Spenard
(Figure 2-2).  The intake for this station was only 17 feet  from the curb, 100
ft. upstream from a traffic light on a four-lane one-way street, and the chart
record suggested that at least some of the high  values recorded  might be due
to instrument error.  But even so, the comparison with the "downtown" station
was startling.  7th and C, for instance, had one day in  January  1979 with an 8-
hour average of 9 ppm or more, the highest 8-hour average for  the month being
12.4 ppm.  Benson and Spenard had 13 violation days, with a maximum  8-hour
average of 20 ppm and 3 days with 8-hour averages exceeding  15 ppm.   A third
site was installed as a check in fall 1979.  This site,  the  Garden Site, is
located in a church parking lot in a residential area.  The  nearest  traffic
light is several blocks away, and most streets in the immediate  area are so
lightly travelled that few stop signs are even present.   Maximum 8-hour averages
and the number of violation days normally exceed those at 7th  and C,  the down-
town station.  A fourth station, also residential in character,  was  added in
the fall of 1980.  This station, variously known as Jewel Lake or Sand Lake,
shows CO levels near those at 7th and C.

    At the time our study began in 1980 the only thing known about the meteoro-
logical conditions leading to the CO problem was that CO levels  increased as
temperatures dropped (Hoyles, 1980) and wind speeds were low  (usually 3 m
sec"-"- or less).  At least four possiblities existed based on the initial  two
stations:  a street-canyon effect when winds blew along  the major streets; a
Fairbanks-type ground-based radiative inversion, a  modified Los  Angeles-type
situation with cold air from the Alaskan interior being  overridden by warm air
from the Pacific Ocean, or a fumigating lapse rate  (one  resulting from heating
from below allowing an elevated pollutant layer  to  mix down to the surface)
when cold air from Interior Alaska flowed over the  incompletely  frozen waters
of Knik Arm north of Anchorage before entering the  city.  Nothing was known of
the synoptic situation associated with CO episodes, and  there  was no  indication
of whether forecasting was even possible.  Nor was  the distribution  over the
city well understood.


RESEARCH PLAN

     Our approach to these problems was as follows:

1.  Fairbanks Forecasts

     This was a relatively minor part of the study.  Fairbanks morning fore-
casts (6 am) were compared with the levels actually reached that day.  We
placed primary emphasis on cases involving forecast and  observed levels on
opposite sides of the 15 ppm level, and secondary emphasis on  cases  at the 9
ppm level.  The results are in Chapter 3.

-------
2.  Anchorage Meteorology

     Comparison of daily and hourly values at different stations allowed us to
determine almost at once that the problem was city-wide, ruling out  the street-
canyon hypothesis.  Airport soundings and hourly observations  were used to
determine cloud cover, wind speed and'direction, temperatures  and  lapse rates
at the airport; temperatures and winds from the monitoring stations  allowed us
to estimate how these factors changed over the greater Anchorage area.   Thermal
infrared satellite photos also provided qualitative information on how  tempera-
ture and lapse rate varied over the area.  Historical  maps of  meteorological
conditions allowed us to isolate the types of synoptic events  most often associ-
ated with high CO levels.  The state and the Municipality of Anchorage  provided
data on CO levels at the monitoring sites and at some  short-term bag  sampler
sites.  Attempts were made to use a Tethersonde (TM) system to monitor  the
variation of lapse rate across Anchorage late in the winter of 1981-82  and
again in 1982-83; and while due to instrument problems only one good  record
was obtained, that record agrees well with deductions  from other data.   Some
informal forecasting of high CO episodes was attempted, with reasonable success,
during the 1982-83 season.  Some relatively simple changes in  the application
of standard dispersion models were suggested, and CO levels, normalized to
time of day and day of week, were compared with the modified stability  classes.
The results of all these observations are given in Chapters 4  and 5.

3.  Fai rbanks mixing heights

     The Tethersonde system was operated in Fairbanks  in March 1981  and in
December 1981.  Three good sets of ascents showing both background and  city
lapse  rates were  obtained in December.  The results are given  in Chapter 5.

     Overall results of the study have been promising.  Although more low-
level soundings are needed in Anchorage due to to the  equipment breakdown
in Jan. 1983, it  is now possible to state what is going on physically to
trap CO in the area, to recommend a forecasting scheme, and to indicate
those parts of the city where the potential for pollution is highest.  Some
comments can also be made about the situation in the Cook Inlet basin as a
whole.

-------
                                   SECTION 3

                             FAIRBANKS CO FORECASTS
FORECASTING METHOD
     Forecasting of CO levels 1n the Fairbanks area has  two  goals.   One  is  to
provide a warning of dangerously high CO levels to individuals  with  health
problems which make them particularly susceptible to this  pollutant,  so  that
they can avoid going into high pollutant areas.  The other is to  discourage
unnecessary driving and encourage preheating of cars which must be driven when
high CO levels are expected, in an effort to keep the CO concentrations  from
reaching health-threatening levels.  To aid in this effort,  Borough  bus  fares
are suspended whenever an air pollution alert (based on  a combination of
observed and forecast levels) is called.  Forecasts are  made twice a day,
between 6 and 7 am and around 3 pm.  In this study, we concentrated  on the
morning forecast, as this is the one which has the greatest  potential  for
modifying people's choice of transit modes or shopping plans.   It should be
pointed out that in general the 3 pm forecast is the more accurate of the two.

     The forecasting method is fairly simple.  An objective, persistance-based
forecast is prepared by assuming that the ratio of the highest  8-hour CO level
for the day to the most current 8-hour CO level available will  be the same  as
the same ratio for the previous day.  Thus if the mean CO level from 10  pm
last night to 6 am this morning were 3 ppm, the mean level from 10 pm night
before last to 6 am yesterday were 2 ppm, and yesterday's  maximum 8-hour level
were 8 ppm, the maximum 8-hour CO level forecast (made at  6  am) for  today
would be obtained from the equation

                     X = 8
                     I   7

and the preliminary forecast would be 12 ppm.  This forecast is then  modified
subjectively by comparison with a dispersion forecast issued by the  Fairbanks
office of the National Weather Service.  The dispersion  forecast  consists of a
categorization of the anticipated dispersion (excellent-good-fair-poor-very
poor), comments on how the dispersion is expected to change  through  the  day, a
copy of the 2 am airport sounding, and usually some comments on a,iy  expected
weather changes.  Continuing with the example of our preliminary  forecast
above, suppose the dispersion forecast showed clear skies  and a substantial
inversion at 2 am, but by 6 am clouds were visible in t^e west  and snow  with
10 knot winds was forecast by noon.  The dispersion forecast itself  might be
"dispersion fair to poor becoming good by afternoon".  The forecast  would
almost certainly be modified downward from the initial 12 ppm,  and as released
would probably be somewhere in the 6-8 ppm range.

     The 6 am forecast of the maximum 8-hour CO level anticipated has been

-------
compared with the maximum 8-hour level actually reached for the three winters
1979-80, 1980-81 and 1982-83.  On the basis of the first two winters, we  sug-
gested that comparison of winds at the airport with winds measured  downtown  be
considered as part of the forecast.  This was not possible during  1981-82, and
unfortunately was not implemented in 1982-83 either.  The 1982-83 winter  had
7 days with 15 ppm or greater compared with 5 for the  two winters previously
studied, so it was added to the earlier set to expand  the number of missed
forecasts available.


FORECAST PERFORMANCE

     Figures 3-1, 3-2 and 3-3 are scatter diagrams of  early morning forecasts
against observed 8-hour maximum CO levels for the three winters  studied..  Not
quite half of the 1982-83 forecasts were within > 2ppm of the observed  level.
Although there is some positive correlation (on the order of .6) between  fore-
cast and observed levels, the prediction of alert levels is poor.   Of 12  days
over the three winters with 8-hour maximum CO levels of 15 ppm or more, not  one
was forecast to exceed 15 ppm, and 5 were not even forecast to exceed 9 ppm.
These 12 days are listed in Table 3-1, along with 15 days with observed levels
of 9 ppm or more which exceeded forecast levels by at  least 5 ppm.   An  alert was
in effect Dec. 21,  1982, but this was due to a forecast of 17 ppm  (maximum
observed 14.5 ppm) the previous afternoon.  There is a psychological  resistance
to forecasting 15 ppm or greater without strong indications, due to the alert
mechanism and the cost of free transit fares.  However, case 9 is the only one
which was even a near miss.

     Separate plots were made for 1982-83 of the preliminary persistance  fore-
cast against observed levels (Figure 3-4) and the Weather Service dispersion
forecast against observed levels (Figure 3-5).  Neither was as accurate as the
Borough's forecast as issued.  There is roughly a 2 ppm bias in  the 1982-83
Borough forecasts, but forecasts that year were by a new forecaster.   Regular
comparison of his own forecasts with reality by the forecaster might  be of
help here.  The cases in Table 3-1 are identified by number or letter in  Figures
3-1 through 3-5, so that the contributions to the errors on individual cases
can be  analyzed.  Starting out with Category A, all four of the  1982-83 cases
were underforecast by persistence and then modified downward.  Category A
cases are shown in detail in Figure 3-6.

     Case 1 was a clear case of a dispersion forecast  which verified with good
winds at the airport but which on the basis of CO levels probably had very
light winds downtown.  This is one of the cases which  led us to  recommend that
wind downtown be used in forecasting.
     In Case 2, forecast dispersion was fair becoming  good based on cloud
cover.  The rise in CO correlated with a brief break in the cloud  cover around
8 am and ceilings above 10,000 feet after that time, but the critical factor
appeared to be a combination of strong warm air advection below 900 mb with
radiative cooling at the surface.  The result was an inversion of  15.3° over
                                       10

-------
2
CL
Q.
_f
HI

LJJ


O
O

DC
X

co   12
2
X
 5
 >-
 Q
 HI
 >
 cc
 LU
 CO
 CD
 O
     20



     18



     16



     14
i  10
     8



     6



     4



     2
           1979-1980
                         20
                      ,8
                       i   i   i   i
                                           17
                                    13
                                                1
 7AM
      0     2     4     6     8     10   12

     FORECAST DAILY MAXIMUM 8 HR CO LEVEL.PPf
Figure 3-1.  7:00 am forecast of day's highest 8-hour CO level against maximum
8-hour level observed the same day, 1979-1980.  Numbered points are in Table

1.
                              11

-------
 QL
 Q.
  fk
 -J
 UJ
   16
 w,  14
O
O

cc
X
00
    12
    10
 x  8
     6



     4



     2
Q
ID

oc
LU
CO
m
O   0
        "J   i—i—i—i—i—i—r


         1980-1981         1
                   18,
                    19'
                           I	1
                                         i—i—r/©rr
     0     2     46     8    10    12    14

  6:30AM FORECAST DAILY MAXIMUM 8 HR CO
                                                     16    18

                                                    LEVEL, PPM
Figure 3-2.  6:30 am forecast of day's highest 8-hour CO level against maximum
8-hour level observed the same day,  1980-1981.  Numbered points are  in Table 1
                                12

-------
OL20
Q.

-J
LJJ
>
LLJ
 o
 o
 LU
 O
 <
 cc
 LJJ
QC
I

CO

2
ID
2

X
<
  15
    10
 Q
 LU
 >
 cc
 111
 CO
 CD
 O
                   T
        1982-1983
                          3
                          4
                                   0

H2
                         /-
                               /
                                 /
                                   /
                                   /
                                     /
                                       /
                 •  \/.
         /
          *y
          s .
         /•e |
          •  •
              1 4"   '
              is**    /
    "0            5            10

   6AM FORECAST DAILY MAXIMUM 8
                   15

               CO LEVEL,
Figure 3-3  6-00 am forecast of day's highest 8-hour CO level against maximum
8-hour level observed the same day, 1982-1983.  Numbered points are in Table

1; I's are ice fog days.
                             13

-------
Table 1:  Poorly forecast cases examined in detail.   Arrows  show forecast
changes with time; the underlined dispersion category was  used  for Figure  3-5

Category A:  observed > 15 ppm, forecast < 9 ppm

                                    observed    forecast     dispersion     persistence
1
2
3
4
5
Jan
Dec
Dec
Feb
Feb
Category
6
7
8
9
10
11
12
Dec
Feb
Nov
Dec
Dec
Jan
Feb
Category
13
14
15
16
17
18
Nov
Jan
Feb
Nov
Nov
29
8
15
4
22
B
17
13
19
23
21
21
23
C
20
29
8
12
17
Dec 29
1981
1982
1982
1983
1983
observed >
2 ppm
1979
1980
1980
1980
1982
1983
1983
observed > 9 ppm
1979
1980
1980
1980
1980
1980
15.7
15.2
17.8
17.6
15.6
15 pm, forecast
20.4
16.0
15.5
16.8
16.1
15.0
18.3
and more than 5
11.6
11.6
9.9
12.3
13.1
12.1
8.5
8 £ + G 9
8 F + P_ 10
8 P -> G + P (13,4)
5 F 8
> 9 ppm, < 15 ppm and low by more than
12.5
9.5
10
14.5
10 F-P + _P 19
12 P 22
10 F * G 28
ppm above forecast
5.5
3.5
4.5
7
5
5.5
                                       14

-------
Category C cent.

19
20
21
22
23
24
25
26
27

Jan 15 1981
Dec 13 1982
Dec 20 1982
Jan 14 1983
Jan 20 1983
Jan 25 1983
Jan 27 1983
Jan 31 1983
Feb 3 1983
Category D - extreme
and forecasts low by
a
b
c
d
e
f
g
h
1
j
Dec 27 1982
Nov 12 1982
Nov 9 1982
Dec 30 1982
Feb 7 1983
Nov 24 1982
Dec 22 1982
Jan 4 1983
Dec 16 1982
Nov 17 1982
observed
11.9
13.4
14.5
11.1
11.3
13.3
14
12.7
9.9
devl atlons on either
more than 5 ppm
10.6
9.9
2.7
4.0
4.2
3.5
9.7
7.0
9.5
10.4
forecast
5.5
7
7
3
4
5
8
4
4
dispersion
7
10
2
6
2
10
12
7
11
12
dispersion persistant

F-G 10
F 7
G-Ex 10
F 4
F 10
P * F f P 10
G 15
P_ -- G 7.5
or persistence forecast,
G-Ex 19
G-Ex 15
Ex + F-P 1.5
F •*• P 18
F + G + P_ 2
G 13*
F-P 22
F + P_ 24
F > P_ 25
F-P ' 26
*Note on forecast sheet that lock change on building necessitated  estimating  data,
Use of actual data gives 11.
                                       15

-------
                      '   '  '   I  I    '
                        6    8    10    12    14    16   18   20    22   24   26
                           AM  PERSISTANCE FORECAST CO. PPM
  Figure 3-4.   6:00  am persistance forecast of day's highest 8-hour  CO  level
  against the  highest  8-hour  CO  level actually observed the same  day,  1982-1983.
  Numbered and lettered points match those in Figure 3-3.
the lowest 113 m of the atmosphere  (13.5°C/100m) at 2 pm, compared with 3°
over the same height (2.6°C/100 m)  at  2  am.  Low level warm air advection was
a critical factor in several  other  cases,  and  needs more attention in dispersion
forecasting.

     Case 3 was again cloudy,  with  high  CO levels when the ceiling rose.  This
is another case where airport soundings  and winds show reasonably good dispersion,
and differences between airport and downtown conditions are likely.

     In the two remaining cases, time  of day was a critical factor.  The noon
solar elevation 1n February  1s high enough to  allow substantial solar heating
                                    16

-------
£.\J
1 18
a.
LU 16
LU
1
0 14
0
QC
I 12
CO
•5
D 10
2
X
1 8
>•
2 6

• • • h|
I
1
- * • • I -
• TRiO
JrB _
— • .
rf : d
: t «c
• •
•
• ~
,1111111
Ex G-Ex G F-G F F-P P VP
                DISPERSION FORECAST, AM

Figure  3-5.  Morning dispersion forecast against maximum 8-hour level  of CO
observed the same day, 1982-1983.  Dispersion plotted 1s the  lowest jf.ocecas-t
for the day; I.e., a forecast of good becoming poor 1s plotted as po'(\r.  I's
are 1ce fog cases.
                                    17

-------
                 CASE 1
     10 8  9  8   CLOUD COVER
     16 U 2828+   CEILING
                  WIND SPEED
                                     1526UNL  CEILING
                                     488  WIND SPEED
  10 10 8  8 10
   U 12 13  13  8
63    43636
   DEC 8,1982
                                              CLOUD COVER
                                                 CEILING
                                               WIND SPEED
                                              CLOUD COVER
                                                 CEILING
                                              WIND  SPEED
                                              WIND SPEED
                          4   8  12  16  20
                         TIME, HOUR OF DAY

Figure 3-6.   Individual  plots of cases with  observed 8-hour CO levels 15 ppm
or more forecast at less than 9 ppm.  Solid  line 1-hour average CO  dashed
line temperature.  Cloud cover  1s In tenths, celling 1n thousands of feet,
and wlndspeed 1n knots,  following NOAA Weather Service usage.
                                    18

-------
with a break in the inversion, and the observed warming from 2 am to 2 pm on
February 4 (Case 4) was just about enough to break the surface inversion
observed at 2 am.  However, sunset on that day was about 3:45 pm, so a ground
inversion had become reestablished by the time of the evening traffic peak.
In Case 5 skies were clear all day and the coincidence of the- rapid cooling
and inversion development near sunset with the evening peak of traffic was
even sharper.  Given the depth and intensity of the early morning inversion
(19°C over 450 m, or 4.2°C/100 m) and the clear skies, the fair dispersion
forecast seems a little weak, and in fact a 10°C inversion over the lowest 250
m  (4°C/100 m) was still present at 2 pm.  Development of an additional steep
ground inversion after 4 pm, just before the rush traffic, was undoubtedly
responsible for the sharp maximum from 4-7 pm.  Case 12, the following day,
had even higher CO levels.  It was forecast higher than Case 5 primarily
because of similarity to the previous day's conditions, and in the face of a
dispersion forecast of "fair becoming good late this afternoon", with 5-15 mph
winds.  The winds materialized just too late to offset the effect of the
traffic peak while the inversion was building after sunset.  This coincidence
of dusk with peak traffic in February is a risk factor not previously
recognized.

     Of the cases with 15 ppm or more with 9-14.9 ppm forecast, Case 6 follows
the pattern of  1, Case 7 is similar to Cases 4, 5 and 12, Case 8 resembles
Case 2, with over 8°C warming at 950 mb between 2 am and 2 pm, and 9 is a
classic clear-sky, low-wind situation which was relatively well forecast (14.5
ppm forecast, 16.8 pm observed).  Case 10 was overforecast on the persistance
forecast, then  modified downward to little more than half the persistance
level  on the basis of a dispersion forecast of "fair to poor becoming poor
tonight" -- clear and light winds in the Christmas shopping period.  This
episode was caught by an alert based on a high forecast the previous afternoon.
Case 11 was again a classic clear and calm situation, with a steep ground
inversion which  hardly changed 1n overall strength through the day.

     Two of the  category C cases -- 22 and 26-- are worth discussing, as are
Cases  a and b.   The lowest CO levels in situations forecast to have poor or
very poor dispersion are also of interest.  Cases 22, 26, a and b had CO levels
above  9 ppm with good to excellent or good dispersion forecasts.  Cases a and
b  were forecast  relatively well by the Borough.  Both were cases of high over-
cast conditions  with ground inversions, and the overcasts probably led to the
good-excellent  dispersion forecasts.  Case 22, however, has to be classed as a
very bad dispersion forecast.  The 2 am sounding showed a 12°C inversion in  the
bottom 50 m  (24°C/100 m), with warm air advection in the lowest 50 mb in pro-
gress.  Case 26 appears to fall into the same category as Case l--good winds at
the airport but  probably not downtown--with the added complication of low level
warm advection.

     The majority of the poor dispersion cases with CO levels below 6 ppm were
1ce fog situations, and the Borough forecasters were  aware enough of the tend-
ency for CO levels to fall off during ice fog that these cases were fairly well
forecast  (See  I's on Figure 3-3).  There are several  possible  reasons for this


                                       19

-------
association.  The cold-start contribution weakens during ice fog,  as a vehicle
which is not pre-warmed will not generally start at ice fog temperatures.
The radiative effects of ice fog tend to improve vertical  mixing (Bowling,
1970) and the difficulty and discomfort of driving at -40°C tends  to dis-
courage unnecessary trips.

     There were three cases with poor or fair to poor dispersion forecasts
without ice fog and with CO levels less than 6 ppm.  Case  c had  fair to  poor
evening dispersion forecast on the basis of snow expected  to stop  by evening;
in fact snow was falling and there was a complete overcast through 2 am  the
following day.  The remaining two Cases, d and e, could be described as
inverses of Case 2, with strong cold air advection in the  lowest 100 mb  of the
atmosphere, and overcast skies which in both cases had been expected to  become
partly cloudy by evening.  Case f, the worst low forecast  by the Borough, was
not a real forecast, as a change in locks on the building  barred the forecaster
from data on current CO levels.


RECOMMENDATIONS

     The general impression received from these analyses and from  discussions
with Borough and Weather Service personnel is that the single most critical
factor in improving forecasts is improved communications between Borough and
Weather Service.  At the present time, the Weather Service forecasters have no
access to CO levels, either as immediate forecast input or to check  how  accurate
their forecasts were.  Figure 3-5 is the first such comparison made.  Borough
forecasters often receive forecasts which are no more than the current sounding
and a one or two word dispersion forecast.

     Several specific meteorological situations do seem to be important  in CO
violations.  One, recognized quite early, involves wind speeds adequate  for
dispersion at the airport with near-stagnation conditions  downtown.   This pat-
tern was documented in Bowling and Benson (1978) although  its significance for
high CO levels was not recognized at that time.  Low level  (below  900 mb)
advection of warm or cold air may also be critical, with warm air  advection
enhancing and cold air advection inhibiting ground-based inversions.  A  high
or thin cloud cover may, as 1n Anchorage (see next chapter) allow  the formation
of a sufficiently steep inversion to promote unacceptably  high CO  levels if
winds are very low.  Finally, as illustrated by the February sunset  cases, the
exact timing of changes 1n dispersion is critical.  An increase  in wind  speed
at 4 pm, just before the traffic peak, will have a far different effect  from a
similar increase four hours later.  This kind of exact timing is an  extremely
difficult forecasting problem, but changes linked with solar elevation angle
are more tractable.  The potential for rush hour CO peaks  just after sunset in
February, for Instance, needs to be recognized.  (This problem will  have a
shift 1n timing as the new time zone comes into effect this year,  and may become
more of problem for Anchorage 1n late January and early February).  The  effects
of high CO Inputs during the Christmas shopping season need also to be kept in
mind.


                                       20

-------
                                  SECTION 4

                                  ANCHORAGE
INTRODUCTION
     Anchorage, looked at on a small  scale,  is  on  a  peninsula jutting into the
head of a major inlet of the Pacific  Ocean - Cook  Inlet.  The area is noted
for being windy and stormy, and scarcely seems  a likely candidate for a major
air pollution problem.  However, if the larger  area  is considered, it becomes
apparent that the whole Inner half of Cook Inlet,  together with the lower
parts of the Matanuska and Susitna Valleys,  form a single large basin (Figure
4.1).  Furthermore, it is apparent from thermal  infrared satellite imagery
that the land portions of the basin floor are often  substantially colder than
the surrounding uplands - in other words, inversions are common (Figure 4.2).
Finally, the precipitation anomalies  of Anchorage  tend to parallel those of
the Interior rather than those of much closer coastal stations located directly
on the shores of the Gulf of Alaska - an additional  indication that Anchorage
does not have an exposed, fully maritime climate,  but is sheltered by the
Chugach and Kenai Ranges,  Nevertheless, it  was  not  immediately clear why CO
levels in Anchorage were so high.


OBSERVED CARBON MONOXIDE LEVELS

     Our first step toward solving the problem  was to compare the CO levels as
a function of time at the three sites then available.  Figure 4.3 shows hourly
CO levels, together with temperatures and airport  lapse rates, for the three
stations operating in February 1980.   The similarity of behavior at the three
stations strongly suggests that the problem  is  area-wide rather than due to
such quasi-local causes as wind alignment parallel to heavily travelled streets.
The fact that Garden Site, a residential area with no concentrated local sources,
frequently had CO levels above the downtown  site,  7th and C, and on occasion
even exceeding the levels at Benson and Spenard  (a high traffic site) was
startling.  This behavior is one of the major pieces of evidence for the wind-
shadow model to be presented later.

     Comparison of the CO levels directly with  meteorological data was compli-
cated by the fact that CO levels depend on sources as well as on dispersion.
Figure 4-4 gives some idea of how current traffic  counts vary with time of day.
In an attempt to remove the regular diurnal  variation in source strength and
meteorology, we calculated for each site a series  of hourly  normal CO levels.
Hourly means and standard deviations  were calculated separately for 1) each month,
2) each Friday, Saturday, Sunday and  3) each Monday  through  Thursday.  Normalized
CO values were then constructed using the formula
                                       21

-------
Figure 4-1.  Composite Landsat photo of the Cook  Inlet area.
                          22

-------
Figure 4-2.  NOAA-4 thermal Infrared Image of the Cook Inlet area.  Anchorage
Is at ANC, and light tones are cold.
                                   23

-------
    Ill
    QC
    r
    Ul   -
IU
5
0
-5
10

15
?0
jX\
«!'•".. 	
• •"•'

.

•

A
_-.

~.^ii

    O m
    *->
      Q.
      Q.
      o"
      o
IU
5
0
-S
•_/

N

^^- 	

"^.^

^x

\^




>N 	

	 -'*'

~~~~^-^
-

              10
             SUN.
 11     12
MON.  TUES.
 13    14
WED.  THUR.
 15
FRI.
 16
SAT.
 17
SUN.
 18
MON.
 19 '
TUES.
                                    FEBRUARY 1980
 Figure 4-3.   Lower curves:   7th  &  C  (solid)  Benson and Spenard  (dashed) and
 Garden Site  (dotted)  hourly CO values.  Middle curve: 100-m temperature
 minus surface temperature (100-m inversion strength) at Anchorage airport.
 Upper curves:  Airport (dot-dashed),  Benson  and Spenard (dashed) and Garden
 Site (dotted) surface temperatures.
     1500
         ° 4  1  2 IT 5  6  7 8  9 10 11 12 13  14 15  16 17 18 19 20 21 22 2324
                                HOUR  OF DAY
Figure  4-4.   Traffic  counts  (averaged  over  5-8  days)  for Benson and Spenard
(solid  line)  and Sand Lake (dashed line).   Based  on December  1982 and January
1983  data,
                                       24

-------
      CO
        norm =
               coobs  "  COjnean  (month, hour, day of week)
                          (month,  hour, day of week)
 where C0norni Is the normalized  CO  level, C00bs is the measured CO
 level, COmean 1s the mean  and OQO  the standard deviation of all CO
 observations taken at the  same  month, hour, and day of the week (Monday through
 Thursday considered as the same day) as C00bs-  Tne resulting values of
 Cnorm f°r the same time period  as  Figure 4.3 are shown 1n Figure 4,5,
 together with wind speeds  and cloud cover measured at the airport.
 METEOROLOGY

      High normalized CO levels  tend to be associated with clear skies or fog,
 low wind speeds,  substantially  lower temperatures at. the measuring sites than
 at the airport,  and Inversions  or  Isothermal lapse rates in the lowest 100 m
 of the airport sounding.    The  presence of a well developed inversion throughout
 the area on the  morning of  Saturday, 16 February can be inferred from the
 thermal  infrared Image shown  in Figure 4.6, which shows the low-lying peninsula
          -2
              10
             SUN.
 11     12     13
WON.  TUE8. WED. THUR.

                 FEBRUARY 1980
 15
FRI.
 16
SAT.
 17    18    19
SUN.  MOM. TUES.
Figure 4-5.  Lower curves:   Normalized  hourly CO values for same stations and
time periods as Figure 4-3.  Middle curve:   cloud cover; solid area cloud,
open area fog.  Top curve:  wiridspeed.
                                     25

-------
Figure 4-6.  NOAA-6  thermal Infrared image,   16 Feb 1980.  8:39 am.
                                    26

-------
on which the city is situated to be distinctly colder (lighter in color)  than
the neighboring mountains.  (The difference in temperature  between  the  airport
and the measuring sites seen in Figure 4.3 argues against the direct  applicability
of the Anchorage airport soundings, although the combination  of  inversions  at
the airport with substantial (5°C or more) temperature differences  between  the
airport and the monitoring sites does appear to correlate well with CO  levels.)
By 2 pm Saturday the airport lapse rate was superadiabatic  (2.7°C/100 m)  for
the first 67 m, capped by a 2°C/100 m inversion over the next  127 m.  Ground
temperatures at the measuring stations, however, were all below  that  at the
top of the superadi abatic layer at the airport.  Although the  airport inversion
reestablished itself overnight, the 2 pm Sunday sounding showed  an  even more
superadiabatic lapse rate - 4.6°C/100 m over the first 46 m -  capped  by a
1.5°C/100 m inversion to 280 m.  Although a ground inversion  might  have persisted
at Benson and Spenard (which was 4.4°C colder than the airport)  CO  levels were
near normal for Sunday.  CO levels remained normal on Monday  in  spite of  clear
skies and relatively light winds at the airport.  Satellite photography and
the airport sounding clearly show a normal lapse rate, with the  peninsula
distinctly darker  (warmer) than the mountains (Figure 4.7).

     This general pattern - clear skies, light winds, substantial temperature
differences between the airport and the CO monitoring sites and  evidence  for
ground inversions from satellite imagery and/or airport soundings - was associ-
ated with the majority of the violation days examined. Locally  developing
radiative inversions of the same type responsible for the Fairbanks problem
are thus strongly suggested as a cause of the Anchorage problem  as well.
Neither the inversions nor the calms are as intense as in_Fairbanks, which
agrees with the fact that Anchorage (1980 population 174431)  has  a problem
similar in magnitude to that of Fairbanks (1980 population  22645; Fairbanks
North Star Borough population 53983).  The pattern is consistent with an  anti-
cyclonic core situation  as is seen in Fairbanks, but examination of weather
maps for the majority of the Anchorage CO violations  over the  last four years
has shown no truly anticyclonic episodes.  The overwhelming majority of cases
have occurred with Anchorage in the southern fringe of a high  pressure system,,
often with a fairly substantial north-south pressure  gradient.   Geostrophic
surface winds invariably have an easterly component,  and can  at  times be  quite
strong.  The observed winds at the Anchorage airport, however, are  normally
quite light.  The weather map during the most severe episode  of  1982-83,  with
violations at all four stations, is shown in Figure 4.8.  A similar map for 2
pm the afternoon of Friday. February 16, 1980 appears as Figure  4.9; this is
one of the most strongly anticyclonic cases observed.

     A second type of weather situation was noted on two occasions: 26  December
1979 and 7 February 1983.  In both cases the core of a dissipating  low  pressure
system was situated over Anchorage.  The 1979 case occurred in the  late after-
noon and evening; the sky was overcast, ceiling 4,000 - 6,000  feet  and  airport
winds were 4-7 knots.  The 2 pm sounding showed a 2°C/100 m inversion.  Obser-
vations from Merril Field, fairly close to Garden Site, indicate winds  of 2-3
knots until about 5 pm, then 4-8 knots; Merril Field also reported  that the
                                       27

-------
Figure 4-7.  TIROS-N thermal  Infrared Image,  18 Feb 1980.  1:35 pm.
                                    28

-------
    Figure 4-8.
Weather map,  2 pm AST
3 Dec. 1982.
                                            Figure 4-9.
Weather map,  2 pm AST
16 Feb. 1980.
moon was visible through the clouds around 9-10 pm.  The meteorological  data
from the monitoring  stations themselves show wind speeds generally 2-3 mph.
The 1983 case had a  major spike coinciding with the 8 am rush hour; although
hourly levels reached 21 ppm at Garden Site, they exceeded 6 ppm for only  4
hours.  The sky was  overcast except for 9/10 cloud with less than 3,000 ft.
ceiling at 8 am.  Winds were light, with a calm at 8 am.


TRAFFIC AND SEASONALITY

     The mean values calculated in the course of normalizing the CO levels
contain a good deal  of information in their own right.   Figure 4-10 shows  the
February mean hourly values for the period of record for each station.   Note
that the years over  which the means were calculated differ for the different
stations; consequently the absolute values for the various stations should not
be compared.   However, the behavior with time of day and to some extent with
time of week is worth studying.

     All  of the stations show what appears to be the classic commuting  peaks
at 8 am and 4-6 pm,  the morning peak being advanced and the evening one delayed
somewhat at the residential  location,  Garden Site.  Evening levels tend to be
higher on weekends at Benson and Spenard (which is closer to being an entertain-
ment district).  Morning peaks on Saturday and Sunday are much lower and delayed
Garden Site is in a  church parking lot, but Sunday values show little effect
of this location.

     The midday dip  in CO levels is due to a combination of factors—not only
is traffic presumed  to be lighter (though  Figure 4-4 does not indicate  this),
solar heating increases dispersion.  Because Anchorage  is so far north, however,
                                        29

-------
        DC
        <
        ID
        DC
        00
        LU
        U_
        Q_
        Q_

        CO
        _1
        LU
        >
        LU
        _J

        o
        o
        z
        <
        LU
        DC
        3
        O
                  	Week   	Sat
                  	Fri      	Sun
                    ,-•"  %'7TH &  C
                        1977-1981
5

4

3
                I   I   T
              GARDEN  SITE,-
               1980-1981 '
                               V
              . \
            •' .• "':~'^ xx'
             i   i
                           J	I
9

8

7

6

5

4

3

2

1
    T  I   I   I   I   I   I   I  I   I   I
             A  BENSON &
             /i  SPENARD

                      I   I   I   I   I   I  I   I   I
             0     4     8    12    16   20   24

                        HOUR OF DAY


Figure 4-10.  February mean hourly values of CO for the  three stations with
the longest periods of punched data.
                              30

-------
solar heating is relatively weak 1n December and early January.  This allows
traffic and solar effects to be disentangled to some extent, as can be seen  in
Figure 4-11.  The sharp rise at 8 am is still present on weekdays,  but at  all
but the residential site the CO levels now remain high through  the  day.  The
evening peak appears as an additional rise from the level  reached in the morning.
Christmas shopping may be having some effect, but comparison with January
means (which are Influenced more by solar heating than December's)  suggests
that the major cause of the dip between the peaks in February is  meteorological.
Garden Site does retain separate morning and evening peaks in December,  but
the minimum between them is now a distinct lunch-hour low  point rather than  a
broad period of low CO levels.


VERTICAL TEMPERATURE STRUCTURE

     Anchorage is over 3 1/2° south of Fairbanks and we had initially  assumed
that midday solar heating would be sufficient to break the nocturnal  inversion
for at least an hour or so a day.  However, this is not the case.   Attempts  to
modify standard dispersion-typing schemes to fit Anchorage first  suggested
that inversions probably persisted through the day in midwinter,  and this was
confirmed by the December mean hourly CO values and by December soundings at
the airport.  Figure 4-12 shows a number of soundings taken at  1  am and  1 pm
during the period from 15-24 Dec. 1982.  Although the strongest inversions are
for 1 am, the systematic difference in inversion strength  between 1  am and 1
pm 1s minor and there are cases (e.g., Dec. 14) where a 1  pm sounding  has a
stronger ground inversion than either of its flanking 1 am soundings.

     Unfortunately, airport soundings are even less appropriate measures of
lapse rates at the monitoring sites in Anchorage than is the case in  Fairbanks.
Anchorage has not only a heat Island, but a major neat source in  the tidal
waters of Cook Inlet as well.  Although 1ce pans cover the water  surface to a
variable extent,  they are thin compared with pack  Ice farther north and there
Is normally at least some open water between the pans.  The  result  1s  that
when winds  speeds  are low enough  that turbulent mixing 1s  minor, the CO measuring
sites are generally colder -- by  as much as 10°C --  than the  airport, which is
nearer the  coast.   This 1s true even though the Anchorage  heat  island  is strong
enough t.o be visible on some NOAA satellite Images,  e.g.,  Figure 4-2.

     If temperatures Inland at some fixed elevation were the  same as those
measured at the alport, the ground temperatures at the measuring  sites could
be used to estimate the lapse rates at the same points. Heat island  effects
would undoubtedly give some near-surface mixing -- probably  more  than  in Fairbanks
(next chapter) -- but the overall  Inversion strength would be correct.   Figure
4-13a shows the resulting sounding for Dec. 25, 1982 at Benson  and  Spenard 1f
the 100-m temperature above the surface 1s assumed to be the  same as  at  the
a1 rport.
                                       31

-------
          	Week   	Sat
          	Fri     	Sun
cc
LJJ
m
UJ
o
HI
Q
Q.
Q.
CO
UJ
LU
O
o
z
UJ
cc
o
              6
              5
              4
              3
              2
              1
              0
              8
              7
              6
              5
              4
              3
              2
              1
              11
              10
              9
              8
              7
              6
              5
              4
              3
              2
              1
           i—i—i—i—i—r
       \ \
        ,	•'  7TH & C
            1977-1980
i   i  i   i  i   i
        •'/    GARDEN SITE
        ./      1979-1980
          i   i   i  i   i  i
                         BENSON
                         SPENARD
                         1978-1980
                                j	I
           4    8    12   16   20   24
                HOUR OF DAY
Figure 4-11.  December mean hourly values of CO for the same  three
Figure 4-10.
                                               Cations as
                      32

-------
         Dec 14, 1om, 7, -5.0  	Dec 15, 1pm, 7, -2.8
        • Dec 14, 1pm, 7, -3.3  	Dec 16, 1am, 2, -5.0
        • Dec 15, lorn, 2, -3.3  	 Dec 16, 1pm, 6, -2.8
          CO
          o
          UJ
                                 — Dec 17, 1am, 1, -3.3	Dec 18, 1pm, 9, -7.2
                                 •—Dec 17, 1pm, 8, -3.3  	Dec r, 1:,n, 3, -7.8
                                 — Dec 18,'1am, 5, -7.2  	 Dec I?, 1frn, 7, -7.2
      — 0«c 20,
      •^ Dec 20,
      — Dee 21,
       2  .500
lorn, 3,-7.2  —•—Dec 21, 1pm, 11,-8.9  	Dec 23, 1am, 5,-12.2	Dec 24, 1pm, 12,-12.8
1pm, 13,-6.7  	Dec 22, 1am, 5,-12.2  ——Dec 23, 1pm, 12,-14.4  	Dec 25, 1am, 7,-12.B
lam, 3,-11.7  	Dec 22, 1pm, 10,-12.2	Dec 24, tarn, 3,-13.9  	 Dec 25, 1pm, 5,-11.7
                12   -10  -8   -6   -4  -2
                            0    -14  -12  -10   -8   -6   -4
                             (1982)              T,°C
-2   0
Figure 4-12.   Airport  soundings for a  high CO episode  1n December  1982.   Heavy
lines are 1 pm soundings.   The first number  after  each  day and time is  the
corresponding  hourly mean CO  (ppm)  at  Benson  and Spenard;  the  second  is  the
surface  temperature at  Benson and  Spenard.
                                               33

-------
     500
   E 400

   UJ
   >

   m 200
   LJ
   X
      100
        0
      500
    .. 400
  >  300
  o
  00
  <

  h-  200
  LJ
      100
        0
                                                        A
               -12    -10    -8     -6    -4    -2

                               T,°C
0
Figure 4-13.  Possible soundings at Benson and Spenard for 1  pm Dec. 25, 1982.
See text for explanation.
                                  34

-------
      However,  air  coming  from  across  the Chugach  range -- the norr^l direction
 during  a  pollution episode  --  will  be moving downward as well as y.-astward.
 The  exact  path followed will vary with the  episode  (Figure 4-14), but the
 uppermost  path --  a straight line from the  range  westward to the ground at the
 airport -- will  give the  maximum vertical air motion between the n.-asuring
 sites and  the  airport.  Air at the  airport  will have subsided about 1UO m
 (Wz) since Benson  and Spenard, and  200 m relative to that over the Garden
 Site.  If  the  air  subsides  as  a uniform layer, and  has a lapse rate (- dT/dz)
 of q, the  temperature difference at a given height  more than 145 n above
 sea  level  between  Benson  and Spenard  and the airport will be Wz(^   Q),
 where Q is the adiabatic  lapse rate of l°C/100m.  Figure 4-13b si ^ s the
 resulting  December 25 sounding at Benson and Spenard.  The real case is
 probably  somewhere between  the extremes.

      In an effort  to resolve the problem, we attempted to use the Tethersonde
 system  to  obtain a chain  of low-level  soundings along Tudor Road.  By the
 time problems  which had  shown  up while using the  system in Fairbanks were
 resolved,  it was too late to use the  system in Anchorage in the late winter
 of 1982.   When we  attempted to try  the same experiment the following year,
 the signal acquisition  circuits on  the ground station went out, requiring
 continual  manual reaquisition  of the  instrument package signal.  By the time
 the instrument was sent  to  Colorado for repairs and returned, the season was
 again over.

      We did obtain one  good ascent  at the Bus Barn  (See Figure 2-2) on January
 14, 1983.   This ascent,  together with the flanking  airport soundings, is shown
 in Figure 4-15.  Although it  confirms our suggestion that inversions become
Figure 4-14.  Possible paths for air flow over the Chugach Range.
                                       35

-------
oi.rpnger  away  from the coast, the warmth of the temperatures  above 25 m is not
easily  explained.  These measurements urgently need  repetition.

THE WIND  FIELD

     The  wind  field in Anchorage, like that in Fairbanks,  is  complex and variable.
Figure  4-16  shows wind directions for three stations  -- the airport, 7th end C,
      200
    -  100
   x
   CD
   UJ
   IE
         -17
                                    \
                    JAN. 14 2:OOA.M.\
                        AIRPORT
                                  	T~
                                                  \
                                                   \
                                  JAN. 14  2:00 P.M.'
                                      AIRPORT
           -16
Figure 4-15.   Comparison of an inland Tethersonde  sounding made at the Bus
Barn (B on  Figure  2-2) with flanking soundings  from the airport.
  2m
  h- O
  Eco
  Q LU
    UJ
  Q f£
  •z. »
  _ UJ
        360
       300
        200
100
          0
                  '
                          fl

                        \£•••''  '•
                      ...V  •    •
                                  DAY, QEC. 77
 Figure 4-16.  Comparison of  wind  directions measured simultaneously  at the
 airport (solid line), 7th and C  (dashed  line) and Tudor and Lake Otis  (dotted
 line).                              36

-------
and an early site at Tudor and Lake Otis -- over a 3-day period with high  CO
levels at 7th and C.  The airport anemometer had a starting speed around 3
mph, compared with 1 mph for the other two stations,  so there are considerably
more missing data at the airport.  Nevertheless, the  general  tendency  is clear
-- northeast winds at the station nearest the mountains swinging through south-
east winds downtown to southwest, winds at the airport.   According to Anchorage
residents the opposite situation, with south winds near the mountains  and
north winds at the airport, also occurs; and we have  observed situations with
fairly uniform wind directions, and with brisk winds  in the western  part of
town while calm conditions prevailed near the mountains.  Additional complica-
tions are introduced by topography.  Fairbanks is on  a  depositional  river
plain with very minor relief.  Anchorage, however, is on an originally smooth
sedimentary plain which was uplifted by glacial rebound and then dissected,
and the local relief, even away from the mountain front, is far greater  than
observed in Fairbanks.  Gravity drainage is correspondingly more complex.

      In view of the uncertainties regarding the spatial variability  of both
inversion strength and winds across Anchorage, we strongly recommend a measure-
ment  program aimed at obtaining and interpreting data on both.  A reasonable
system would involve  a 15-m tower at each CO monitoring site plus at least four
additional towers along an east-west transect -- probably along Tudor  Road --
and  one or two towers in topographically interesting  areas.  Wind and  temperature
measurements should be made at 2, 8 and 15 meters.  A modern data logging
system would allow direct  readout of hourly mean wind directions and speeds,
standard deviation of direction and possibly periodic variations induced by
gravity waves.  This  is particularly important as the existing micrometeorolog-
ical  stations at the  monitoring sites are being removed.


THE  LARGER BASIN:  INNER COOK  INLET

      The CO  problem appears to be local and confined  to the immediate  vicinity
of Anchorage.  However, the larger basin is subject to  inversions, as  seen in
Figures 4-2  and 4-6,  and both  shores of the inner half  of Cook Inlet,  as well
as large portions of  the lower Matanuska and Susitna  valleys, must all be
considered part of the same air drainage.  An unplanned demonstration  of that
fact  occurred in the  spring of 1983, when moderate clearing operations  (500
acres) were carried out near Point Woronzoff, across  Cook Inlet from Anchorage.
Slash burning from the clearing resulted in a very substantial  smoke problem
in Anchorage (Anonymous, 1983; Ryan, 1983).  According  to my own observations,
at around 5 pm 20 May 1983, the plume was crossing the  Parks Highway as  a  well
defined layer near the Knik River bridge, at the head of the arm of  Cook Inlet
north of Anchorage.   In the valleys inland of the bridge the smoke was eddying
visibly down to the ground in what appeared to be a fumigation type  episode.
The  lower atmospheric layers had in all probability remained stable  and  inhibited
mixing while crossing the  cold waters of Cook Inlet,  then been destabilized  by
heating as they moved over the land.

      Superadiabatic lapse  rates near the ground with  strong capping  inversions
starting at 50-300 m have  been observed at Anchorage  in midwinter.  These


                                       37

-------
episodes are associated with northerly to westerly surface winds  and are almost
certainly due to heating of the lowest layers of cold continental  air flowing
over the incompletely frozen waters of Cook Inlet.  Of two such  episodes in
January 1980, one was associated with severe CO pollution (1-hour  levels above
15 ppm at Benson and Spenard and Garden Site) and the other with 0-1 ppm.  The
major difference appears to be that in the first episode, skies were clear and
the temperatures at the monitoring stations were almost 10°C cooler  than at
the airport -- i.e., it appears that a ground inversion was developing  very
rapidly as the air moved inland or that intend stat-ions were not influenced  by
Inlet-modified air.  The capping inversion was relatively we'ak.  In  the  second
case, skies were cloudy, the CO monitoring sites were only 2-3°C cooler  than-
the airport, and the capping inversion was quite strong--la30/100,  m  from 280
m to 400 m.  The extreme superadiabatic lapse rate on 17 "Feb." 1980,  which was
capped at less than 50 m and accompanied by average CO levels, has already
been mentioned.  Thus the elevated inversion appears to have a negligible
influence on CO concentrations.  However, this situation could cause an  elevated
plume from across the Inlet to mix down to the ground at Anchorage.

     These two cases suggest that potential dispersion problems may  exist
throughout the year for the larger basin.  Emissions on the west shore  of Cook
Inlet must be considered in terms of their influence on the entire basin.
This should certainly be taken into consideration in any discussion  of  industrial
development, as well as more extensive clearing planned for Point  Woronzoff.


CONCLUSIONS

     The major meteorological situation associated with high CO levels  in
Anchorage, Alaska is a ground-based radiative inversion with wind  speeds  low
enough that turbulence is minor.  Surface temperature data and very  limited
sounding data both suggest that inversions become more intense and winds  become
lower away from the coastal areas and toward the mountain front.   Conditions
probably improve in the foothills, but no measurements are available.  The
very high CO levels observed at Garden Site, with far less source  intensity
than Benson and Spenard, tend to verify this:  Garden Site is the  only CO
monitoring site between the Chugach Range and A Street.  The tendency for
pollution episodes to occur with geostrophic flow from the direction  of  the
mountains strongly suggests that the shielding effect of the mountains may be
very important in the development of the low wind speeds observed  with high  CO
levels.

     On the basis of this study the area most vulnerable to pollution from
local sources (though not necessarily with the highest present levels)  is that
between the Seward Highway and the foothills of the Chugach Mountains.   Partic-
ular care needs to be taken that source levels do not increase in  this  area,
especially in the vicinity of Providence Hospital.

     Forecasting of CO in Anchorage might be possible in a very  limited  way  --
e.g., very low, moderate or moderately high probability of violation or  alert
levels.  Such forecasting would be based on Weather Service capability  and


                                       38

-------
willingness to forecast wind and cloud cover for the entire  peninsular  area,
rather than just the airport, when forecast geostrophic  winds  were  easterly.
An additional requirement would be access at the forecasting location to  real-
time CO data from a good monitoring station.  Although  our own forecasting
attempts as part of the 1983 field season did not pick  up  any  violations  at
the established sites, we did manage to identify in  advance  the  day which
provided the Tethersonde ascent with a substantial  inversion,  as well as  get
the principal investigator to Anchorage during a period  with several high
spikes of CO.

     It is.-also worth pointing out that Anchorage,  like  Fairbanks,  has  a  penod
in midwinter when ground-based "nocturnal" inversions can  form regardless of
time of day.  As a  result, the mixing volume available  for daytime  automotive
emissions is far less than it would be 20° further south.

     Details of both the vertical temperature structure  and  the  wind field
through the city are still unclear and need further study.   However, the  funda-
mental cause of the Anchorage problem, like that of Fairbanks, appears  to be
the short days and  lew sun angle in winter.
                                       39

-------
                                SECTION 5

                      MIXING HEIGHTS AND MODELING
INTRODUCTION
    Modeling of air pollution levels is important both for estimating
the effect of changing the pattern of sources and for determining critical
meteorological factors.  Most existing dispersion models were designed  for
situations where high pollution levels are due to high emissions  with only
moderately poor dispersion, and cannot be applied if wind speed or vertical
dispersion is extremely low.  Unfortunately the high-latitude pollution
situation is generally one in which moderate quantities of pollutants are
emitted into an atmosphere in which dispersion is extremely poor.  In
particular, vertical dispersion is normally much worse than the models  can
cope with.  (Low wind speeds and the complexity of wind directions in the
horizontal can also cause problems with standard models, but this is  too
large a problem to address in detail here.)

    As it is still necessary to estimate the effects of new sources or
changes intended to alleviate the effects of existing sources,  standard
models are, however, applied.  Since the worst cases the models can handle
are generally or the order of inversions around 1 to 2°C/100 m  or mixing
heights around 100 m, while observed inversion strengths in Fairbanks may
easily be 10 to 30°C/100 m, it is hardly surprising that the predicted
pollution levels do not agree with those observed.  This is usually
handled by "tuning" the models - adjusting parameters until  the model gives
the observed pollution level.  Unfortunately this procedure cannot  be used
to obtain a model  which responds correctly to a major change in source
distri bution.

    The remainder of this chapter is concerned with measurements  aimed
at obtaining the physical parameters affecting vertical  dispersion  in
the Fairbanks and Anchorage areas, and suggestions of ways in which
existing models might be modified to allow more accurate computations
of pollutant levels in high latitudes.


EFFECTS OF "TUNING" MODELS

    Two simple cases should suffice to show some of the pitfalls  of
using heavily tuned models to evaluate the results of changing  source
fields.  First, consider an area with good mixing up to 50 m, with an
intense inversion at 50-m and an initial source which is primarily an
area source.  A model which can handle a 100-m mixing height as a worst
case can be tuned to handle this situation fairly well  by using a 100-m
mixing height and doubling the emission rate.  But what happens 1f the
                                    40

-------
tuned model is used to evaluate the impact of a proposed  point  source
with an effective stack height of 75-m?  In reality,  the  strong inversion
at 50-m would allow very little mixing to ground level.   The modeled  100-m
inversion, however, would give complete trapping of  a doubled pollutant
input.

    A second tuning problem has to do with the practice of  linking
horizontal and vertical dispersion parameters in most models.  This is
examined in some detail in Appendix 1, and will  be considered here only
briefly.  Going back to our area-source city-, suppose we  now have a model
with horizontal and vertical dispersion parameters linked so that very
low vertical dispersion implied a very uniform wind  direction.   In our
real city, however, the variability of wind direction begins to increase
as vertical dispersion becomes very poor.  If measurements  are  made
within the area covered by the area source, wind direction  and  its
variability are not important, and the model  can be  tuned to handle
the area source even though the variability of the wind direction is
totally  unrealistic.  However, an elevated  point source cannot be
correctly handled in this case.  If realistic vertical dispersion is
used, the horizontal spread of the modeled average plume  will be a few
degrees, while the horizontal  spread of the physical  plume  (averaged
over several hours) will more likely be tens  of degrees.  This  in turn
will produce an order of magnitude error in the plume concentration.
If the actual horizontal variability is used  to set  the dispersion and
the source strength is increased to compensate for the unrealistic high
vertical dispersion this produces, the degree to which the  plume mixes
down to the ground will be incorrectly modeled.

     Given that most air pollution episodes at high  latitudes do in fact
occurwhen vertical dispersion  is very poor (F or  G stability or worsT)
while  winds are light  and very  variable in direction,  it  is  absolutely
essential  that  any  model  used to evaluate  the effects  of a change in
source  distribution  be  capable  of  handling   separate  horizontal and
vertical dispersion parameters.   Furthermore,  both   horizontal  and
vertical dispersion  used in such models should  correspond  as closely as
possible to those actually occurringFinally it  must be recognized that
the meteorology corresponding to that of the  worst case with  a  change in
sources may differ from that for the worst case  with  existing sources.


MEASUREMENTS OF FAIRBANKS MIXING HEIGHTS

   It has been known for many  years that Fairbanks ground-based inver-
sions, measured outside the downtown area, are among  the  strongest in
the world (Benson, 1965).  However, there has been little or no
information on the details of  how the near-surface lapse  rate is modi-
fied by the city heat island,  although Bowling and Benson (1978) specu-
lated that the mixing height in the core area was  around  60  m.  The State
                                    41

-------
of Alaska made a Tethersonde system available  to  us  for  this  study,
and several pairs of ascents were made at  Creamer's  Field,  upwind  of
Fairbanks, and at 5th and Lacy, a block south  of  the CO  monitoring station
(Figure 5-1).  The three best sets of ascents  were carried  out in
December 1981, covering three of the five  days that  month with 8-hour
CO levels over 12 ppm, including both days over 13 ppm.
Fiqur° 5-1.  Main road net in Fairbanks.   CF (Creamer's  Field)  is site
             of background ascent;  D (downtown)  is  the  city  ascent
             location.

-------
    Hourly CO data around the time of the ascents are shown  in  Figure
5-2, and the ascents themselves are shown in Figures  5-3  through  5-5.
If the isothermal part of the city sounding is  taken  to indicate  the
mixing height, then the measured mixing heights were  30 m on  15 December,
10 m on 22 December and 6 m on 23 December.  While these  values do  not
correlate well with the CO levels observed at the time  of ascent, they
do agree with the direction of change in the CO level  observed  at that
time.  The 30-m mixing height was measured as hourly  mean CO  dropped
from 13.5 to 6.5, the 10-m height had CO holding nearly steady  from
15 to 14 ppm, and the 6-m height corresponded to an increase  from 8 to
12.5 ppm.

    The mixing heights on the ascents for 22 and 23 December  could  also
be defined as the heights where the city and background soundings meet,
in which case the mixing heights are somewhere  between  15 and 30 m  for
22 December and close to 50-m for 23 December.   However,  these  portions
of the soundings include substantial inversions - up  to 15°C/100 m.  One
explanation for these warmed but still  inverted segments  of the city
soundings is that they represent partial mixing due to  warm polluted
air rising in contact with building walls coupled with  slow sinking of
the intervening  relatively clear air.  However, the upper part  of the
warming on 23 December, as well as the cooling  from 45  -  95 m on 22
December, could be due to gravity waves in a strong inversion.  In  both
cases, the observed temperature difference from background to city  could
be accounted for by 15-m vertical motion.  Gravity waves  of at  least
this amplitude have been documented in the Fairbanks  area (Holmgren et al.,
1975).  The apparent warming between 30 and 50  m on 23  December could
also represent a power plant plume, as the city power plant would have
been upwind at that time and place.  There are  few buildings  in Fairbanks
exceeding 10 to  15 m in height, and the few taller buildings  (up to 35 m)
are clustered northwest of the measurement site.

    Another method of estimating mixing height  under  very stable condi-
tions is to Identify the height of the low-level wind maximum (Arya, 1981).
Again, the data suggest 30 m or less.  In particular, there is  a definite
maximum of 1.4 m s"^ at 29 m on 23 December. December  22 shows a maxi-
mum wind speed of .4 m s"1 between 3 and 6 m, with wind speeds  from 13 to
25 m being below the instrument threshold, again suggesting very little
mixing above 6 m on this ascent.

    The explanation offered here is that complete turnover and  mixing,
even in the city heat island, extended only to  6 to 30  m, the exact depth
depending on the detailed structure of the lowest portion of  the back-
ground Inversion and on the time the air had spent in the city.  Partial
mixing due to warm air rising along building sides and  wind-generated
turbulence probably moved some pollutants up to 30 to 40  m, regardless
of the overall mixing depth.  It is unlikely that any substantial
vertical  mixing of pollutants above that level  occurred.
                                    43

-------
       a.
       Q.
        •k
       O
       o
     15

     10

      5

      0
                       Background
                        /  /City
                  DEC. 15
i   i   i  i   i   i   i  i   i
       Q.
       Q_
        •t
       O
       o
      15

      10

       5

       0
                       Background
                          /City
                 Lr
                  DEC. 22
   i   i  i   i   i   ii   i
       2  20

       £  15

       O  10

       0   5
             Background
              /  /City
                            DEC. 23
                        i   i
              10    12   14    16   18
                      HOURS
Figure
5-2.  CO levels near the times of the three sets of ascents,
                       44

-------







METERS








12U
110


90
80

70
60
50

40

30
20
10
0
i i i
•
DECEMBER 15, 1981 •
•
. SURFACE WINDS <.6 M s71 *
210° TO 240° E OF N
A CITY /
• RURAL A
•
A
-
A*
A
•
A
^
*•
*
f
, (
f
JKJi.

f^^-jtl^ji^
,fii:>',.'.'SS,-:!J', '*'•)'.'> •\;V.\'
i-'j- si}V/o;'';,'-'i;v>U'''-'- •'-•>
.
OK -90 -15 -10
                    TEMPERATURE °C
Figure 5-3.  Background and downtown soundings, 15  December 1981



                               45

-------
120




110




100



 90



 80
   70
CO
cc
111

£  60
   50
   40
   30
   20
   10
         DECEMBER 22, 1981
        SURFACE WINDS <.5 M s71

        65° TO 100° E OF N
          * CITY

          • RURAL
    -30
                                                   A •
                -25          -20

                TEMPERATURE °C
-15
   Figure 5-4.  Background and downtown soundings, 22 December 1981.
             Different shade patterns show city site warmer or colder

             than rural site.
                            46

-------
  120


  110


  100


   90


   80


   70
CO
QC
£  60
UJ

   50


   40


   30


   20


   10
                                               •
       DECEMBER 23, 1981
    SURFACE WINDS < .3 M s71
    155° TO 200° E OF N
    30 M WINDS UP TO 1.7 M s71
    310° TO 330°  E OF N


      * CITY

      • RURAL
                        .--•'^'"•Hr
                        A^v'-'vv1:*
                       .K'M'-'';;7V-''Vv-'A
-25
    +
                                         •A
                                         ,\ ' '
                                         /o-
                                           -10
               -20         -15

               TEMPERATURE °C

Figure 5-5.  Background and downtown soundings, 23 December 1981.
                         47

-------
    These measurements do not represent worst-case conditions.   All  of  the
three cases had background inversions of 5°C or more  in  the  first  35  m
(15°C/100 m) and fairly high CO levels, but stronger  background  inversions
(Bowling, 1967), and higher CO levels are known to occur in  the  Fairbanks
area.  Observations along the Steese Expressway in the southeast part of
Fairbanks show clear layering of smoke at heights  of  10-12 m to  be common
in winter.  All things considered, 10 meters should probably be  considered
a maximum mixing height for a worst-case simulation of pollution from
near-surface sources in the Fairbanks area, and worst-case wind  speeds
should be modeled at .5 m s"' or less.  The variability  of surface wind
directions is well illustrated in Figures 5-3 through 5-5 -  the  range of
values in each figure occurred within about 5 minutes.   Worst case analyses
for elevated sources should assume a mixing height just  above the  effective
stack height.

    Further discussion of the effect of city heating  on  mixing height is
found in Appendix 2.


ANCHORAGE MIXING HEIGHTS

    Understanding of the appropriate parameters for modeling the Anchorage
situation 1s not as satisfactory as 1s the case for Fairbanks.   This  is
in part due to the fact that much more was known about Fairbanks than
Anchorage at the beginning of the study, 1n part due  to  equipment  problems,
and in part due to the contribution of local topography, including Cook
Inlet, to the complexity of the Anchorage situation.  The tower  measure-
ments of winds and Inversions recommended in Section  4 are badly needed
as Input for modeling.  Interim worst case values  should be  based  on the
Fairbanks situation - partial mixing to building heights (much higher in
Anchorage than Fairbanks), complete mixing to 10 m, wind speeds  .5 to 1 m
sec"1.  Further Tethersonde measurements are urgently needed in  Anchorage.
CONCLUSIONS

    Input values for modeling worst-case dispersion  in either  Fairbanks
or Anchorage should not exceed 10-m mixing height or .5 to  1 m sec"1 winds.
HoMzontal'and vertical dispersion must be separated,  as the worst-case
situation generally Involves A horizontal  and G vertical  dispersion  classes,
The wind fields 1n both cities are complex, and modeling over  the  full
area should be done with models capable of handling  differences 1n wind
vectors of 180° over distances of as little as 20 m  vertically or  a  few
hundred meters horizontally.
                                    48

-------
                               SECTION 6

                        SUMMARY AND CONCLUSIONS
CONCLUSIONS
     The Anchorage CO problem, like that 1n Fairbanks, is due to nocturnal
inversion conditions persisting through the hours of maximum automotive
emissions.  The high CO emissions characteristic of cold starts undoubtedly
add to the problem, and encouraging the preheating of engines at tempera-
tures above -20°C could have beneficial effects in reducing emissions.
Emission control strategies in use at lower latitudes would also help by
reducing emissions from warm vehicles and possibly reducing the number of
cars operating.  In contrast to the common correlation of nocturnal
inversions with anticyclonic conditions (as is the case in Fairbanks),
Anchorage CO episodes occur due to shielding of the city by the adjacent
mountain front.  All but two CO episodes were found to occur with a
substantial pressure gradient between an anticyclone to the north and a
cyclone to the south.  Blockage of the resulting easterly geostrophic
winds by the Chugach Range (which forms an abrupt front east-southeast
of Anchorage) can lead to stagnation conditions over the city.   There
is good evidence that inversions are better developed near the  mountain
front than along the coast, and that winds are often lighter near the
mountains as well.  Thus the potential for high CO levels, if not the
actual current levels, may be worst just west of the mountain front.

     The two exceptional cases occurred with dissipating low pressure systems
located directly over Anchorage.  Thin or high clouds with essentially  zero
pressure gradients seem to be responsible for the poor dispersion in  these
cases.

     Coupling of horizontal and vertical  dispersion within the  model  is
known to cause problems with the use of Gaussian plume-based models in
nocturnal inversion conditions, and this problem extends to any use of
such models at high latitudes.  The "split-sigma" approach of Sagendorf
and Dickson (1974) is essential for model  applications in Anchorage and
Fairbanks.  In addition, vertical stability categories must be  based  on
solar elevation angle rather than time after sunrise.  There is good
evidence that nocturnal conditions may persist as much as 3 hours after
sunrise at high latitudes.

     Mixing heights were measured directly in downtown Fairbanks, and
found to be as low as 6 meters.  Ten meters is probably a generous esti-
mate for a worst-case mixing height in Fairbanks, and should be used  in
Anchorage until better measurements can be made.

     Any serious modeling of the effect of changing source distributions
in the Anchorage area must take account of the complex and, at  this point,
very poorly understood wind field.  A program of meteorological tower
measurements, 1n conjunction with further tethered balloon observations,
1s urgently needed 1n Anchorage.

                                   49

-------
     There 1s evidence that fumigation-type conditions can occur summer or
winter within the larger upper Cook Inlet basin.  This needs  to be  considered
if development of the north and west shores of Cook Inlet  is  contemplated,
as the entire inner basin may act as a single airshed.

     The Fairbanks forecasting scheme, although showing some  skill  and
offering a definite improvement over either pure meteorological  dispersion
or pure persistence-based forecasts, is not forecasting the most severe CO
episodes.  Meteorological features associated with 8-hour  CO levels in
excess of 15 ppm include relatively high winds at the airport with  calm
conditions in downtown Fairbanks, low-level warm air advection, and
detailed coincidence of poor dispersion conditions with emission peaks.
Several striking examples of the last condition occurred in February,
when the nocturnal inversion, although broken during midday,  was re-
established by the time of the 5 pm emissions peak.  The single factor
which would probably do most to improve forecasts is better communication
between the Fairbanks North Star Borough CO forecaster and the local
office of the NOAA Weather Service (which prepares the dispersion fore-
casts).  At present the dispersion forecasts are prepared  without benefit
of CO data, and the Weather Service is not getting CO feedback to validate
these forecasts.   The Borough forecaster, on the other hand, receives
dispersion forecasts which vary enormously in completeness - ranging from
a one-word dispersion forecast and a copy of the latest airport sounding
to a detailed discussion of how weather and dispersion are expected to
change through the next 24 hours.  Both parties would benefit by closer
interaction.

     A probability-based forecasting scheme for Anchorage  could probably
be set up with existing data.  Such forecasting would require real-time
access to CO data  (not presently available in Anchorage) and full coopera-
tion from the local NOAA Weather Service Forecasting Center.
 RECOMMENDATIONS

      1.  An  observational micrometeorological program should be carried
 out  1n Anchorage to clarify the wind field and the local variation in
 inversion strengths.  The program should include 15-m towers at each CO
 monitoring site, along an east-west traverse across the city, and at
 least one tower each 1n a drainage channel and on a local ridge.
 Measurements should Include wind and temperature at 2, 8 and 15 m, and
 modern data-logging techniques should be used.

      2.  Any attempt to develop an adequate a1r pollution model for
 Anchorage would require the data from 1, above.  As an interim recommenda-
 tion, a  10-m mixing layer and wind speed of 0.5 m sec"1 should be used for
 worst-case modeling of surface sources.  Models used should separate
 horizontal and vertical dispersion, and solar elevation angle rather than
                                   50

-------
time relative to sunrise or sunset should be used to  estimate vertical
dispersion.

     3.  Communications between the Fairbanks North Star Borough and the
NOAA Weather Service must be strengthened if alert levels of CO are to be
forecast for Fairbanks with any degree of accuracy.   Similar communication
lines 1n Anchorage should be assured as an initial condition of any possible
forecasting effort there.
                                   51

-------
                                 REFERENCES


1.   Anonymous.  Burning brush adds smoke to spring dust.  Anchorage Daily
    News, April 26, 1983, page 1.

2.   Arya, S.P.S.  Parameterizing the height of the stable atmospheric boundary
    layer.  J. of Appl. Met., 20, 1981, pp. 1192-1202.

3.   Benson, Carl S.  Ice fog - low temperature air pollution defined with
    Fairbanks, Alaska as type locality.  Report of the  Geophysical  Institute,
    University of Alaska, UAG R-173, 1965, reissued with revision  in 1970
    as CRREL Research Report 121, Hanover, New Hampshire. 118 p.

4.   Billelo, Michael A.  Survey of arctic and subarctic temperature inversions.
    CRREL Technical Report TR 161, Hanover, New Hampshire, 1966, 35 p.

5.   Bowling, Sue Ann.  A study of synoptic-scale meteorological features
    associated with the occurrence of ice fog in Fairbanks, Alaska.  M.S.
    Thesis, University of Alaska, 1967, 141 p.

6.   Bowling, Sue Ann.  Radiative cooling rates 1n the presence of  ice crystal
    aerosols.  Ph.D. Dissertation, University of Alaska, 1970, 365 p.

7.   Bov;ling, Sue Ann  and Carl S. Benson.  Study of the subarctic  heat island
    at  Fairbanks, Alaska.  EPA  report 600/4-78-027, 1978.  U.S. Environmental
    Protection Agency, Research Triangle Park. 150 p.

8.   Bowling, S.A., Takeshi Ohtake and Carl S. Benson.  Winter pressure
    systems and ice fog in Fairbanks, Alaska.  J. of Appl. Met.,  7, 1968,
    pp.  961-968.

9.   Fahl, C.   Internal atmospheric gravity waves at Fairbanks, Alaska.   M.S.
    Thesis, University of Alaska, 1969, 94 p.

10.  Holmgren,  B., L. Spears, C. Wilson and C. Benson.  Acoustic soundings of
    the  Fairbanks temperature Inversions.  Climate of the Arctic,  G.  Weller
    and  S. A.  Bowling,eds.  Geophysical Institute, University of  Alaska,
    Fairbanks, 1975, pp.  293-306.

11.  Holty, Joseph G.  Air quality in a subarctic community, Fairbanks,
    Alaska.  Arctic 26, 1973, pp. 292-302.

12.  Hoyles, M.  A study of wind patterns in Anchorage,  Alaska that are
    associated with violation of the carbon monoxide standard. Report  from
    the  Alaska Department of Environmental Conservation, 1980.

13.  Jayaweera, K.O.L.F., G. Wendler and T. Ohtake.  Low cloud cover and the
    winter temperature of Fairbanks.  Climate of the Arctic, G. Weller and
    S.A. Bowling, eds.  Geophysical Institute, University of Alaska, Fairbanks,
    1975, pp.  316-322.


                                     52

-------
14.  Leonard, Leroy.  Cold start  automotive  emissions 1n Fairbanks, Alaska.
     Geophysical  Institute, University  of  Alaska, Fairbanks.  Report UAG R-239,
     1975.  132p.

15.  Leonard, L.  E.  Carbon monoxide emissions from moving vehicles in
     Fairbanks, Alaska Vol. 3.  Geophysical  Institute, University of Alaska,
     Fairbanks.  Report UAG R-252,  1977.   45  p.

16.  Ohtake, Takeshi.  Studies  on ice fog.   Geophysical Institute,University
     of Alaska Report UAG R211, reprinted  by  Office of Air Program Pub. No.
     APTD-0626.  U.S. Environmental  Protection Agency, Research Triangle Park,
     NC. 1970. 177 p.

17.  Oliver, Vincent J. and Mildred  B.  Oliver.   Ice fogs in the interior of
     Alaska.  Bull. Am. Meteor. Soc. 30, 1949, pp.  23-26.

18.  Remsberg, Ellis E., James  J. Buglia and  Gerard E. Woodbury.  The nocturnal
     inversion and Its effect on  the dispersion  of carbon monoxide at ground
     level in Hampton, Virginia.  Atmos. Environ. 13, 1979. pp. 443-447.

19.  Ryan, Andy.   Farmland burning  fouls city air.  The Anchorage Times,
     April 28, 1983, p. B-4.

20.  Sagendorf, J. F. and C. R. Dickson.   Diffusion under low windspeed,
     inversion conditions.  NOAA  Air Resource Lab, Idaho Falls, 1974, 93 p.

21.  Weller, Gunter, editor. Ice fog studies in Alaska.  Geophysical
     Institute, University of Alaska, Fairbanks. Report  UAG R-207, 1969, 49 p.

22.  Wendler, G.   Relation entre  la  concentration en oxyde de carbone et las
     conditions meteorologiques dans une communaut£ subarctique.  J.
     Rech. Atmos.  9, 1975, pp.  135-142.

23.  Winchester,  J. W., W. H. Zoller, R. A.  Duce and C. S. Benson.  Lead and
     halogens in  pollution aerosols  and snow  from Fairbanks, Alaska.
     Atmos. Environ. 1, 1967, 105-119.
                                      53

-------
                                  APPENDIX 1
          MODIFICATIONS NECESSARY TO USE STANDARD  DISPERSION MODELS AT
                                HIGH LATITUDES*
     It has  long been recognized that GaussIan-plume models with plume
spreads based on semi-empirical  stability categories do not apply to conditions
of very low  wind speeds (< 1  m sec'1).  Nevertheless, regulatory agencies
may require  application of such  models, even In areas such as  Alaska where
1t 1s recognized that many, 1f not most, violations occur under conditions
which the models cannot handle.   It is the purpose of this contribution to
examine the  approximations behind these models and their associated dispersion
criteria, to point out those that conflict with observed meteorological condi-
tions and to suggest .modifications to these assumptions which  would improve
the utility  of these models under conditions of strong inversions and very
light winds.
     The fundamental physical assumptions behind Gaussian models can be
illustrated by the release at a point x=0, y=0, z=z0 of successive puffs
                                 +
of a pollutant Into a wind field V(x, y, z, t).  For the moment we assume
that the puff density matches that of the surrounding air, and the vertical
                                                                        *
velocity at release =0.  A single puff will follow a trajectory of the V
field, with some spreading due both to molecular diffusion and the fact
that if the initial puff has non-zero dimensions its component subregions
will follow slightly different trajectories.  If the concentration as a
function of x, y, z at a time t after each puff is calculated for a very
large number of puffs and the results averaged, the general appearance
                                                    ~
will be that of a puff traveling with the mean wind V and spreading in the
                                                          •*•  +    •»• >
x, y, and z directions with the turbulent fluctuations of V; v1 « V-V.  If
the coordinate system 1s oriented so that the mean wind direction 1s along
the x axis,  the mean position of the center of the puff at time t will be
*to be published in  Atmospheric Environment
                                    54

-------
given by (vxt, 0, z0) and the magnitude of the spread in the x, y and z
directions will be determined by the distribution of magnitudes of v'x,
Vy' and vz'.  The distribution of pollutant concentration along any axis
from the point (v"xt, 0, z0) is normally assumed to be proportional to
                             	      1   x^-Xi  2
the Gaussian distribution,  /Zno-j exp [-^  —0    ].  Thus far,
there is no obvious reason why the theory should not apply to very small
values of 7X.
     If the release of a pollutant Is continuous, the time-averaged con-
centration at a given point 1s the Integrated total of contributions of puffs
released at all times from ta-»tot = 0.  At this point an approxi-
mation is normally made which Implicitly assumes that vx' < 7X : mixing
in the x direction  1s Ignored on the assumption that the variation in
pollutant concentrations between parcels which might be exchanged along
the x axis is negligible.  However, this assumption does not always hold
under clear-night, radiative-inversion conditions, as indicated by
measured extreme-to-extreme variations in wind direction in excess of
180° (Fahl, 1969).  In this case vx'  > Vx, and non-negligible pollutant
concentrations may occur "upwind" of the source.  This particular
assumption is too deeply rooted in conventional  Gaussian-plume models
to be removed without completely redoing the model', and will  cause
problems with upwind concentration modeling whenever the mean amplitude
of vx'  equals or exceeds Vx.
     Widely-used Gaussian-plume models such as the HINAY and CALINE
models make another assumption which can be expected to cause significant
errors under conditions of low wind speed:  horizontal and vertical dis-
persion are assumed to be under the control of the same dispersion
                                    55

-------
parameters.  The 01 1n the Gaussian distribution above 1s a measure of
how far the plume has spread horizontally (ay) or vertically (az)
and 1s a function of the Vy' (or vz') values and the time since the plume
element left the stack:  effectively, t » x/v~x.  If the o's are tabulated
as functions of distance, then the 
-------
perpendicular to the wind, however, 1s negligibly affected by horizontal
diffusion 1f the target point 1s located within the area or substantially
closer to the line source than the line source length.  The effect of
vertical  stability Is roughly the same for area and point sources.  If a
real  situation has a horizontal  dispersion corresponding to category A
(extremely unstable - 09 > 25°)  and a vertical dispersion corresponding to
category G (extremely stable -  AT/AZ > 1°C/100 m) any attempt to
apply a single stability class in modeling will over-estimate the importance
of point sources relative to area sources.  Such data as are available for
Fairbanks, Alaska suggest that the A-G combination above is by no means
an exaggeration, and in fact area sources appear to dominate the air pollution
distribution in Fairbanks.  Recent paired urban ano rural  measurements there
have documented rural inversions of 10°C/100 m with urban mixing heights of
10 m and urban wind speeds of 0.5 m sec'1, with directional fluctuations over
a range of 40° (Bowling, 1984a).
     An additional assumption in Gaussian models in that the dispersion cate-
gories and Vx are independent of the space and time coordinates.  This is certainly
not true 1n the vertical: 180° wind shears over heights of a few hundred
or even tens of meters are common in Fairbanks (e.g. Holmgren  et al., 1975),
and long-term-shears of similar magnitude can occur over a horizontal  distance
of less than a kilometer (Bowling and Benson, 1978).  The effect of such variation
with z is partially offset by the fact that the lack of momentum transfer necessary
for the development of large vertical wind shear implies also the lack of
vertical  pollutant transfer through the shear, but the horizontal variability
remains a serious problem in modeling.
     The actual assignment of horizontal  and vertical dispersion criteria
at high latitudes 1s also subject to some problems.  Table 1 summarizes
several criteria for Pasquill and Turner stabilities and their relation-
ships and is based on Sagendorf and Dickson (1975) and Gifford (1976).
                                     57

-------
                               TABLE 1
                  DEFINITION OF STABILITY CATEGORIES
Classification
Extremely unstable
Moderately unstable
Slightly unstable
Neutral
Slightly stable
Moderately stable
Extremely stable
Pasqulll
A
B
C
D
E
F
G
Turner
1
2
3
4
5*
6*
7*
Slade:o0 A
(degrees)
25
20
15
10
5
2.5
1.7
T/az(°C/100 m)
-1.9
-1.9 to -1.7
-1.7 to -1.5
-1.5 to -0.5
-0.5 to 1.5
1.5 to 4
>4
*The relationship between these categories and the Pasqulll  categories is
1n some doubt; this relationship 1s the one used Implicitly by Doty and
Holzworth  (1976).

     The standard deviation of wind direction (09) should be used to
obtain ay.  Unfortunately, OQ is not generally available from standard
meteorological data, although Ooppler radar wind printouts may include
09 as a derived number.  Such data as are available from Fairbanks
suggest that OQ values may significantly exceed 25° under near-calm con-
ditions.   It 1s possible that horizontal dispersion at specific sites
might prove to be closely related to a combination of vertical stability
and *1nd speed, but this has not yet been Investigated and any relationship
would probably be site specific.
     Actual lapse rates are rarely available at the times and places where
they are most needed for pollutant modeling, so the vertical stability
category 1s normally estimated from a combination of sky covor and cloud
                                     58

-------
height, wind speed, and solar elevation angle.  Standard schemes (e.g.,
Doty and Holzworth, 1976) also Involve a distinction between day and
night based on time' (one hour after sunrise and one hour before sunset).
Physically, the distinction between  day-unstable and night-stable
conditions lies in whether the absorbed solar radiation  is  greater  than
or less than the net outgoing longwave radiation.  Ignoring long-wave
differences, this means the physical  dependence is on albedo,  solar
elevation and cloud conditions.  As shown by Figure 1, ono  hour after
sunrise corresponds to greatly differing solar elevations at different
latitudes, especially as the latitude exceeds 50°.  Consequently, use of
the standard tables can assign C or even B stability to  cases  with  lapse
rates typical of G stability - at Fairbanks, for instance,  the sun  at
winter solstice is less than 2° above the horizon at solar  noon,  almost
2 hours after sunrise, and the daily temperature maximum is most likely
to occur at midnight.
     In an attempt to correct this problem, we drew up the  following
modification of vertical stability calculation based on  the instructions
given 1n Doty and Holzworth  (1976):
     (1)   If the total cloud cover is 10/10 and the ceiling is less than
2000 m (7000 ft) set the net radiation index » 0 if the  solar elevation
                                       1,
angle 1s greater than 6° and ijqual to - <2 if vthe solar  elevation angle
is less than 6°.  Go to stsp 5.
     (2)   Use Table 2 to determine the insolation class.  (Solar eleva-
tion > 60° gives class number.4).  For any particular latitude and  longi-
tude, a chart can be drawn up giving the solar elevation angle as a function
of tine of day and date, as has been done for Anchorage  (150°W, 61° 10'M)
It  Figure  2.
                                    59

-------
     lOi
  V)
  UJ
  Q
     81
                                                    SO'
  UJ

  CC
  o
  CO
3

2

1
        s\
      67 66 65
                     60
53
50
45
Figure 1.  Dependence on latitude of solar elevation  one  hour after
           sunrise at the winter solstice for high latitudes.   The
           equinox value at Los Angeles 1s about 12°, which  1s also
           the value assumed 1n the practical definition  of  civil
           twilight (1/2 hour after sunset or solar depression of
           6°).
                                  60

-------
                              NOV
6AM
7AM
            DEC
JAN
FEB
MAR
SOLAR  ELEVATION  ANGLES  FOR
    ANCHORAGE,  160*  W,  61« 10'N
                                                BEFORE
                                               SUNSET
                   Figure 2.  Solar elevation angles, sunrise and sunset for Anchorage. Alaska,
                             as a function of time and date.  The conventional day-night
                             boundary for dispersion classification is shown, dashed, for
                             comparison; a working chart would have only the solid lines.

-------
                                  TABLE 2
                    INSOLATION CLASS DETERMINATION
Solar
Bare
Snow
Snow
Elevation
ground
patchy or <
coyer 6" or
Angle

6" deep
more
> 35°
+3
+3
+2
35 21
+2
+2
+2
18
+2
+2
+1
15
+2
+1
N/2
12
+1
N/2
night

6 <
N/2
ni
ni
ght
ght
6

night
ni
ni
ght
ght
     (3)  If the category is "night",  use a net radiation  index  of  -3  if
skies are clear,  -2 if the cloud cover is between 0.1  and 0.4,  and -1  if
the cloud cover is 0.5 or more.   Divide these values  by 2  for  "N/2".   Go to
step 5.
     (4)  For daytime:
          a.  If the cloud cover is less than .5,  the net  radiation index
equals the insolation class.  Go to step 5.
          b.  If the cloud cover is greater than .5,  modify the  insolation
class number by following the steps below:
          (1)  Celling < 2000 m (7000  ft) subtract 2.
          (2)  Celling between 2000 m  and 5000 m (16,000 ft)  subtract  1.
          (3)  Cloud cover « 1,  subtract 1.
          (4)  The net radiation index is equal  to the modified  insolation
class nunber or to 1, whichever 1s greater.
     (5)  Use Table 3 to find the vertical stability  class.
                                      62

-------
                                  TABLE  3
                      DETERMINATION OF STABILITY  CATEGORY
Wind
Speed
m sec"'
0 - .5
1
2
3
3.5
4
5
5.5
6
roph °
0, 1
2, 3
4, 5
6, 7
8
9,10
n
12
>13
Knots
0, 1
2, 3
4, 5
6
7
8, 9
10
n
>12
Net Radiation Index
43210
A
A
A
B
B
B
C
C
C
A
B
B
B
B
C
C
C
D
B
B
C
C
C
C
D
D
D
C
C
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
D
-V2
E
E
D
D
D
D
D
D
D
-1
F
F
E
E
0
D
0
D
D
-lV2
F
F
F
E
E
D
D
D
D
-2
G
G
F
F
E
E
E
D
D
-3
H
G
G
F
E
E
E
D
D
     This scheme 1s essentially identical  to the Doty  and  Holzworth  (1976)  scheme
(hereinafter referred to as DH) for the contiguous  U.S.  in summer  and  all  but
the northern tier of states in winter.   The only differences  are the addition
of a brief transition period between night-stable and  day-unstable regimes,
allowance for the effect of snow cover  on  solar heating,  and  addition  of  one
more stability category, "H", tentatively  equated to AT/AZ of more than
10°C/100m.  At high latitudes in winter, however, the  present scheme and  DH
differ spectacularly.  DH predicts that the time interval  between  the  lines
labeled "1 hour after sunrise" and "1 hour before unset"  in Figure 2 will
have neutral to unstable stability.  Plowed streets and general  grime  put
Anchorage Into the patchy snow category, so the modified procedure allows r\o_
unstable conditions In November, December or January.   Average hourly  CO
values 1n these months do not show a nldday decline beyond the traffic-related
                                      63

-------
dropoff between 8 am and 10 am, while February hourly means show a definite
mid-day minimum 1n CO concentration (Bowling, 1984b).
     Additional comparisons of both schemes have been made with Individual
Anchorage hourly CO levels.  The results are:
     (1)  When observed CO concentrations were normalized to January hourly
values for a 2 1/2 month period 1n 1979-80, all  three periods with CO levels
six standard deviations or more above normal were associated with "H" stability,
as were well over half of the periods with CO concentrations 3c or more above
normal.
     (2)  No tendency toward a mid-day minimum of CO was observed before
February 13, although DH predicted several hours a day of "B" stability.
On February 13 the modified scheme for the first time predicted "B"
stability ("H" stability overnight) and a midday drop in CO became apparent
and persisted through the rest of February.
     (3)  During a 1-day period from 14 through  24 December 1982, inversions
persisted at the airport with no systematic difference in stability between
2 am and 2 pm soundings.  Thus far, no case in which the DH and modified
schemes predict significantly different dispersion conditions has been
found in which the modified scheme has not been  more accurate.   Further
testing is needed, especially for the deep-snow  case, but the superiority
to DH at high latitudes in winter seems clearly  established.  "H" stability,
tentatively equated to AT/^ of more than 10°C/100 m, does not appear
1n most tables; but, with the separation of horizontal and vertical dispersion,
1t should be possible to Incorporate G and H stabilities into existing models.
     In closing, we wish to reemphaslze the importance of separating
horizontal and vertical dispersion parameters rather than tuning a coupled
model against observed conditions.  Tuning can be expected to work only
1f the geometrical properties of the sources remain unchanged.  A model
tuned with area sources dominating which Is based on linked horizontal and
                                     64

-------
vertical  dispersion parameters cannot predict the result of a source change
which involves expansion of an area source to surround a measurement
point or the addition of one or more major point sources or line- sources
parallel to the mean wind.  Use of models for such purposes at high
latitudes requires spl1t-s1gma models.
                                      65

-------
                                REFERENCES
Bowling, S. A. and C. Benson (1978)  Study of the Subarctic heat island at
     Fairbanks, Alaska.  EPA report 600/4-78-027, June 1978,  149 .pp.
Bowling, S. A. (1984a)  Climatology of high latitude air pollution.   Submitted
     to J. of Climate and Appl.  Meteor.
Bowling, S. A. (1984b)  Meteorological factors responsible for high  CO levels
     in Alaskan cities.  U.S. Environmental Protection Agency

Doty 4 Holzworth (1976)  Climatological analysis of Pasquill  stability
     categories on STAR summaries.  NOAA Environmental Data Services, DAS
     Ql 882, D6, 61 pp.
Fahl, C. (1969)  Internal atmospheric gravity waves at Fairbanks,  Alaska.
     Master's Thesis, University of Alaska, 94 pp.
Gifford (1976)  Turbulent diffusion-typing schemes:  a review.  Nuclear
     Safety, V17, No. 1, NTIS DAS (A QC 880 A4).
Holmgren, B., L. Spears, C. Wilson and C. Benson (1975)  Acoustic  soundings
     of the Fairbanks temperature inversions.  Climate of the Arctic, eds.
     Gunter Weller and Sue Ann Bowling, Geophysical Institute, Fairbanks,
     Alaska.
Sagendorf 4 Dickson (1975)  Diffusion under low windspeed, inversion  conditions.
     NOAA Technical Memorandum,  ERL-52, Idaho Falls Air Resources  Laboratory,
     December 1974.
                                     66

-------
                                APPENDIX 2
        THE INFLUENCE OF THE  FORM OF THE TEMPERATURE SOUNDING  ON  THE
               DEPTH OF THE MIXING LAYER PRODUCED OVER A CITY *

  X.  Introduction
       Haste heat (which ultimately Includes almost all of the energy used
  by mankind) 1s In the long term a major limiting factor to energy growth.
  In the short term, middle to high latitude cities 1n winter already release
  as waste heat amounts of energy  comparable to or even greater  than that
  received by these cities from the sun.  The result 1s a  noticeable dis-
  turbance of the natural lapse rate 1n the  vicinity of a  city - a disturbance
  which  may profoundly Influence both  local  temperatures and  the mixing of
  pollutants when the natural  lapse  rate  1s  stable.
      When stable air flows  over a warm  surface,  such  as  a city whose rough-
  ness 1s  comparable with that of Us  surroundings,  the resultant heating can
  be  described  1n two ways: the temperature  near the ground rises,  and a mix-
  Ing layer develops 1n the otherwise  stable  air.  The first effect has been
  expressed 1n  an empirical heat Island formula by Ludwig  (1970)  and 1n the
  theoretical treatment by Summers (1965).  The second effect 1s Important
  for dispersion  of  air pollution and has been studied by Leahey  and Friend
  (1971), among others.  The purpose of this paper Is to point out  the
  Importance of the nature of the background Inversion on the  depth  of the
 mixing layer and the  strength of  the heat  Island developed.   (See  also
 Bowling and Benson, 1978).
 2.   Development of the Model
      Assume that air which Initially  has a  potential  temperature  e(z,0)
 which Increases  with height  (z) 1s  flowing  with  speed v(y) Independent of
 h«1ght  over a  uniform surface.  As we are considering only a  shallow layer
*subm1tted to Atmosphere-Ocean
                                 67

-------
of air (< 1 km) we win use the approximate equation
    do    dT                                                       i
    3z  » 3z + r, where r Is the adlabatlc lapse rate,  1°C (100 m)  ,
                                 dT
the lapse rate being defined as~'dT.   The coordinate along a  trajectory
1s y.  Over some part of the surface, sensible heat 1s  transferred  to  the
air from the surface at a rate q(y) expressed 1n energy per unit area  per
unit time.  The total energy which will have been added to a column with a
unit basal area by the time 1t reaches the coordinate y1  1s
              y1
                                                                         (i)
This energy 1s assumed to heat the air column by development  of  an
adlabatlc lapse rate working up from the base of the sounding (Summers,
1965).  The potential temperature profile at point y1  1s  then assumed to
be
     e(z.y') = e(Z,0)        (0 < z < Z)
                                                                         (2)
     e(z.y') « e(z,0)        (z > Z)
where Z « Z(y') 1s the depth of the mixing layer, and AT  =  9  (Z,0)  -
0(0,0) 1s the Intensity of the heat Island (the amount of rise 1n
surface temperature) at point y'.
     Sensible heating of a unit volume of the air 1s given  by cp pA0,
where p 1s the air density and cp 1s the specific heat of air, so  the
total energy required to develop an adlabatlc layer of depth  Z by  adding
an amount of heat Q(y') 1s
                                 68

-------
     Q(y') •  /  cp P[~e(z,o) - o(z,o)l  dz
             'o      L               J
or, provided Z 1s small enough that the variation 1n p with height may
be neglected,
                       Z
     Q(y')
             "	" "    "                                     (3b)
Hy')             I   \               1
—  «  G  (y1)  «  /    etz.o)  -  e(z,o)   dz
  P               -^o  L                J
Once e(z,0) 1s specified, (3b) may be solved analytically or numerically
to obtain Z(y') and thence AT(y') as functions of G(y') and e(z,0).
     Note that G, which 1s the change 1n temperature Integrated over height,
1s proportional to the total heat energy transferred through the base of the
air column.  The maximum value of G, which has units of length temperature,
will vary with city size, energy use patterns, wind speed and air density.
For Fairbanks, Alaska with population 45,000 and wind speeds of the  order
of 1 m sec'1 or less, G at the city center Is believed to be of the  order
of 200 to 250 m K (meters-degrees Kelvin) (Bowling and Benson, 1978).  Three
sets of low-level soundings outside of Fairbanks and near (but not at) the
city center gave values from 45 m K to 120 m K (Bowling, 1984).
     To demonstrate the effect of the form of e(z,0) on both Z and AT,
three Initial soundings with the same potential  temperature e0 at ground
level and the same value of 6(100,0), differing from 90, at 100 m will
be considered: a constant lapse rate profile, a capping Inversion and a
logarithmic Inversion,
     The temperature-height equations for the three cases and their  approx-
mate potential temperature equivalents are: for the constant lapse rate

     T(z,0) - T0 - yz (Y • lapse rate »  ^ )                            (4

-------
or
     e(z,0) « e0 + ez  (0 - r - Y;  r - adlabatlc lapse rate);        (4b)
for the capping Inversion

     T(z,0) - TO - YIZ                    (0 _< z < zi)
                                                                    (5a)
     T(z,0) - TO - Yizi - Y2(z-zi)        (zi <_ z <_
where Y! Is the lapse rate 1n the weakly stable or n..utral air below
the capping Inversion, Y2 *s the lapse rate In the capping Inversion,
and Z] and 22 are the heights of the base and top of the capping Inversion
The sounding 1s not defined here for z > Z£.  The potential temperature
form 1s
     e(z,C) = e0 + 0-jz                  (0 <_ z <_
                                                                    (5b)
     e(z,0) * e0 +
the e's being defined as 1n (4b).
Finally, the logarithmic Inversion 1s given by
     T(z,0) - TO - rz + ab An (1 + f)                               (6a)
or
     e(z,0) « 60 •«• «b in (1 +"5).                                   (6b)

         d9    dT                                             d9   a
Here a • ~&  • gj + r at z • 0, and b 1s the height at which  
-------
 defined,  I.e.,  there 1$  a  functional  relationship between G and Z.  To
 obtain this relationship for  the  temperature distributions given by (4),
 (5)  and (6), substitute  each  In turn  Into  (3b).  The results are, for
 the  linear case:
or
                 ;     AT « VZ0G
                                                                        (8a, 8b)
For the capping  Inversion
             o
                          z
                                        (Q < I < Z})

                            2
                           l
For Z < 21 , (8a) amd (8b) apply with 0 replaced by 01;  for ZT  <_ Z  _<  22
           Y [
                                                                        (10)
    AT
Finally, for the logarithmic case:

     G • ab  {Z - b in (1  •   '
                                                                        (11)
 3.  Numerical results
     Z has been calculated as a function of G for the  three  types of pro-
files discussed above, using two different Inversion strengths  1n the
altitude range from 0 to 100 m.  The soundings are shown  1n  F1g. 1  and
                                 71

-------
   180


   160


   140


£ 120
UJ

UJ 100


H  80


LJJ  60


    40


    20
      n—i—i—r—i—r
                                                   /  r  i—rn—i—r
                                                  / / 4° C/100 M
                                                  / /  INVERSION
-2
Figure 1.
                2    4    6    8   10   12   14   16

                 TEMPERATURE, °C, WITH T0 - 0
                                                        18   20   22
      Temperature soundings for which the mixing heights and pollution
      potentials In Figures 2 and 3 were  calculated.  Heavy  lines -
      overall Inversion 14°C/100 m; light lines - overall  Inversion
      4°C/100 m.  Solid lines - constant  lapse rate; dashed  lines -
      logarithmic sounding; dotted lines  - capping Inversion.  4°C
      capping Inversion 1s terminated at  100 m; 14°C capping Inversion
      has 3 possible extensions Indicated.
                           72

-------
the numerical values of the coefficients are 1n Table 1.  It Is apparent
from the values of 2 as a function of G shown 1n F1g. 2 that the different
forms of these soundings (which are within observed limits for Fairbanks,
Alaska) are sufficient at times to produce more change 1n Z than a  three-
fold change 1n the temperature difference between 0 and 100 m - the type
of parameter most often used to represent the Inversion strength.
4.  Implications for air pollution potential
     If pollutants are added at the ground In direct proportion to  the  heat
added, and are then mixed uniformly through the adlabatlc layer, the con-
centration of pollutants 1n the mixed layer at any point y1 should  be pro-
portional to G(y')/Z(y').  This quantity, which we will call  the pollution
potential Index (PPI) 1s plotted against G 1n Fig. 3 for each of the sound-
Ings considered.  It must be kept 1n mind that this PPI completely  neglects
variation of a number of f<»ctors that may differ among cities or even with
season (e.g., wind speed, or ratio of pollutant output to heat output,  or
air density).  Nor does 1t consider effects of radiative energy losses
(which have the effect of reducing the energy transferred to  the air and
thus Increasing the pollution/heat ratio) or the Impact of elevated pollu-
tion sources.  To the extent that G Increases with city size  (with  the
square root of the population 1f geometrical similarity 1s maintained)
F1g. 3 does provide an estimate of how pollution would be expected  to
Increase as a dty grows.  This assumes predominantly low-level  pollution
sources and does not allow for major Industrialization, which could change
the ratio of heat Input to pollution Input.
     The logarithmic sounding 1s typical  of regions with very low wind
speeds and clear skies, with radiative cooling dominating outside the
dty.  Such conditions are common 1n sheltered locations at high latitudes.

                                  73

-------
                              TABLE 1
Values of coefficients 1n equations (4) through (6) corresponding to
            overall 0 to 100 m Inversion of 14°C and 4°C
      Sounding
      14° constant Inversion
      14° capping Inversion
      Yl - 0°C/m
      Y2 " 1.40°C/m
Parameters
0  = .15 K/m
01 = 90 m
zi = 90 m
02 - 1.41 K/m
Z2 - 100 m
03   (above
   « 0 K/m (adlabatlc)
   « .01 K/m (Isothermal)
   - .05 K/m (4°/100 m 1nvers1or
      14°logarithmic Inversion
a » .985 K/m
b • 5 m
      4° constant Inversion
01 - .05 K/m
      4° capping Inversion,
      Yl • .007°C/m
      Y2 • .463°C/m
01 » .003 K/m
zi » 90 m
02 - .473 K/m
Z2 • 100 m
     Calculations terminated
     at 100 m
      4°logarithmic Inversion
a - .328°  K/m
b » 5 m
                              74

-------
  en
  o:
  LJ
  h-
  LL|
  o:
  Ld

  5
  o


  X
  u_
  o
  LJ
  X
300
     200
     100
        0   200
                 600      1000      1400     1800

                  G, METERS  DEGREES KELVIN
2200
Figure 2.  Mixing heights calculated for various values of G, the change
          1n temperature Integrated over height, for  the soundings  1n

          Figure 1.  Line code 1s the same as Figure  1.
                                 75

-------
*

N
X
LJ
Q
LJ

O
£L
O
Q.
                                           4°C/100M  INVERSION
                500        1000        1500       2000

                     G,  METERS  DEGREES KELVIN
2500
Figure 3.  Pollution Potential Index (PPI) as a function of G for the
          soundings 1n Figure 1 and 2.  Line code 1s the same as 1n
          Figure  f.
                                   76

-------
For this type of sounding, the equations predict a very fast growth of the
PPI with heat Input, so that even a small settlement may have a substantial
pollution problem, In agreement with observed conditions at high latitudes.
However, the PPI then levels off rapidly, becoming asymptotic to the constant
value ab.  Some values of Z and G giving various fractions of the  maximum PPI
are given 1n Table 2.
     It should be emphasized that the PPI approaches a constant value
only 1n the case where the logarithmic sounding approaches the adiabatic
sounding with height.  A more realistic case would be an approach  to a
stable constant lapse rate.  Mathematically, this would correspond to

      0(z,0) - 6Q + Sz + ab £n (1 + "b)                                (12)
            from which
             2
                                                                      (13)
G • •*£-  + ab  [z - b in (1  + b)j
     Figure 4 compares the PPI's corresponding to (12)  and (6b)  when  e
1s set equal to  .005 (a normal lapse rate at half the adiabatic  strength),
b 1s kept at 5 meters, and a 1s adjusted to maintain a 14° Inversion  between
0 and 100 m.  The PPI for  e(z,0) • 6 Q + -005 z 1s plotted for  reference.
The very rapid Initial rise 1n pollution potential with city size  1s
maintained, but  the PPI continues to Increase for large values of  G.  The
values of 0, a and b used 1n F1g. 4 are reasonable winter values at
Fairbanks, Alaska.  This sounding has a maximum temperature 20°C warmer
than the surface at an altitude of about a kilometer.
     The constant lapse rate case 1s the easiest to handle analytically,
with PPI -/0G/2.  Physically, 1t may be considered a special form  of
the logarithmic case In which b -H-while a • B.  Theoretically,  there
1s no upper limit on pollution Intensity.  In practice, the constant
                                    77

-------
                             TABLE 2
Non-dimensional values of G and Z corresponding to various ratios
    of the pollution potential  Index to Its maximum value, ab
PPI/ab G/ab* Z/b
.25
.50
.75
.90
1.00
.18
1.25
7.06
32.4
•
.73
2.5
9.4
36
«•
                                  78

-------
             14° C LOGARITHMIC ASYMPTOTIC TO
             .5°C/100M NORMAL LAPSE  _
                                         14° C LOGARITHMIC
                                 ASYMPTOTIC TO ADIABATIC J^APSE
               .5°C/100M NORMAL LAPSE RATE
Figure 4.
          250
      500   750   1000  1250   1500  1750  2000  2250 2500

         G, METERS DEGREES KELVIN
Effect on the Pollution Potential  Index of having a logarithmic
sounding approach a stable lapse rate rather than an adlabatlc
one.  Note Vertical scale change from Figure 3.
                                   79

-------
 lapse  rate ts normally limited In vertical  extent,  and the situation
 for very  large values  of G will  approach either the logarithmic or
 the capping case.
     The  capping Inversion 1s  generally  due to  large-scale motion rather
 than to radiative processes.   Advectlon  of  warm air over cooler air 1s
 one possible  causative  mechanism.   Subsidence over  a mixed layer or
 heating of a  stable  air mass from below  are other possible formative
 mechanisms.   (Inversions of this  type are typical of I s Angeles).  With
 this type of  Inversion,  the potential for pollution  1s low for small
 settlements but continues to Increase rapidly for large values of G.
 It 1s  not difficult  to  demonstrate that, to a good  approximation, the
 rate at which the PPI  Increases with G 1s Inversely proportional  to  the
 height of the base of  the capping  Inversion.  The capping Inversion  case
 1s unique In  that a  large enough city may be able to "break" the  Inversion,
 at which  point ground level pollution will  fall off sharply.  Manipulation
 of (9) and (10) shows that for an  Idealized  "square" capping Inversion
 (01 •  0,  (22  - zi) « zi,  and  (z2  - zi)  $2  • I • Inversion strength),
 the PPI at the point of breakthrough 1s  directly proportional to  I,  while
 the value of  G necessary to reach  this breakthrough point 1s equal to Ize«
 5.  Conclusions
     As a general rule,  1n areas  1n which Inversions are normally steepest
 near the  ground and  stability  decreases with height, air pollution will be
a problem even In imall  towns  but  Its severity will  Increase only slightly
with Increased dty  size.   In  areas where stability Increases with height,
small   settlements will  have fewer  problems with air pollution, but the
severity  of pollution will  Increase relatively rapidly with dty  size.
                                   80

-------
Thus,  1f  cllmatologlcal  summaries of  Inversion  frequencies are to be used
as Input  for making  decisions related to  pollution,  they should whenever
possible  be organized to Indicate the predominant Inversion forms as
well as Inversion strengths.
Acknowledgement:  This work was supported by The United States Environmental
Protection Agency Grant No. 80299, and the State of Alaska.
                                         81

-------
                                 REFERENCES


Bowling, S. A. and C. S. Benson, 1978.  Study of the subarctic heat Island
     •t Fairbanks, Alaska.  Environmental Protection Agency Report
     EPA-600/4-78-027, available through National Technical Information
     Service, Springfield, Virginia 22161.

Bowling, S. A., 1984.  Meteorological factors responsible for high CO levels
     1n Alaskan-cities, Environmental Protection Agency Report.

Leahey, D. M. and J. P. Friend, 1971.  A model for predicting the depth
     of the mixing layer over an urban heat Island with applications  to
     New York City. J. Appl. Met.. 10; 1162-1173,

Ludwlg, P., 1970.  Urban air temperatures and their relation to  §xtra-
     urban meteorological measurements.  Papers presented at the symposium
     on survival shelter problems, American Society of Heating,  Refrigerat-
     ing and A1r-Cond1t1on1ng Engineers, Jan. 19-22, San Francisco, pp. 40-45.

Summers, P. W., 1965.  An urban heat Island model:  Its role 1n air pollution
     problemi with applications to Montreal,  Paper presented at the  First
     Canadian Conference on Mlcrometeorology, Toronto,  12-14 April  1965,

-------