-------
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,
------- |