Jun 20, 1995 - Robert D. Hudson and Jae-Hwan Kim. Department of ... Anne M. Thompson. NASA Goddard Space Flight Center, Greenbelt, Maryland.
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 100, NO. D6, PAGES 11,137-11,145, JUNE 20, 1995
On the derivation of tropospheric column ozone from radiances measured by the total ozone mapping spectrometer Robert
D. Hudson
and Jae-Hwan
Kim
Department of Meteorology, University of Maryland, College Park
Anne M. Thompson NASA Goddard Space Flight Center, Greenbelt, Maryland
Abstract. We have developed a new algorithm to retrieve tropospheric column ozone on a daily basis directly from the measured total ozone mapping spectrometer (TOMS) albedos at a spatial resolution of about 50 km. This new algorithm is applied to the retrieval of tropospheric ozone over the region bounded by 20øW and 60øE longitude and 20øSand 0øSlatitude during the 1989 biomass burning seasonwhen tropospheric ozone approaches20% of total ozone. We find that for the conditions of high tropospheric ozone during biomass burning, the archived TOMS data underestimates total ozone over cloud free regions and overestimates ozone over the marine stratocumulus cloud off the west coast of Africa. This error can be as high as 15 Dobson units. Thus previous methods which derive tropospheric ozone from the TOMS archived ozone data are subject to error. Details of the algorithm and an analysis of the expected error are presented. 1.
data
Introduction
Satellite measurements of tropospheric column ozone are necessary for sufficient spatial and temporal ozone resolution over the globe. Fishman and coworkers first used total ozone mapping spectrometer (TOMS) data (the archived grid-T product with 1.25øresolution) to examine the variability of tropospheric column ozone in the tropics, assuming that the stratospheric column ozone field was a smooth function of latitude and longitude. Fishman and coworkers [Fishman, 1988; Fishman et al., 1986, 1990, 1992; Fishman and Larsen, 1987; Watson et al., 1990] then developed the "residual" technique, in which a residual tropospheric column ozone amount is calculated by subtracting a stratospheric column ozone amount, obtained from the Stratospheric Aerosol and Gas Experiment (SAGE) profiles, from the low spatial resolution grid-T data. Time-averaged plots of the tropospheric residual [Fishman et al., 1991] show a plume of high tropospheric ozone extending from the west coast of Africa into the Atlantic during the biomass burning season in southern Africa and South America (June to October). These high levels of ozone have been observed over the continent and ocean by ozonesondes at Natal, Brazil [Logan and Kirchhoff, 1986] and Brazzaville and Ascension Island [Cros et al., 1992; Fishman et al., 1992]. However, when high ozone amounts are present in the lower troposphere, TOMS will underestimate the true amount [Klenk et al., 1982], so that neither TOMS total ozone nor the tropospheric residual technique should be used for tracking tropospheric ozone. Hudson and Kim [1994], Thompson et al. [1993], and Fishman et al. [1987] have pointed out the limitations of using archived TOMS data over low clouds. In particular, analysis of daily TOMS Copyright 1995 by the American Geophysical Union. Paper number 94JD02435. 0148-0227/95/94JD-02435505.00
shows
that
total
ozone
over
marine
stratocumulus
clouds could be overestimated by as much as 15-20 Dobson units (DU). In this paper we present a new algorithm designed to overcome some of these difficulties. Daily tropospheric column ozone maps are retrieved directly from the TOMS measured albedos. These albedos are archived on the highdensity TOMS data tapes (the HDT tapes), and we are able to obtain maps at a spatial resolution of about 50 km square, which is approximately 6 times more dense than the grid-T TOMS product. We have appliedthe techniqueto the tropical easternAtlantic, the adjacentAfrican continent,and the Indian Ocean, during the biomass burning season in 1989. The 1989 season was selected because several published reports have focused on TOMS ozone and the meteorology of this region [Hudson and Kim, 1994; Thompson et al., 1993; Fakhruzzaman et al., 1994; Pickering et al., 1994a, b]. Firstly, we will review the basics of what TOMS actually measures, and how the current algorithm retrieves total ozone (for further details the reader is referred to Klenk et al. [1982] and Dave and Mateer [1967]). Then we will discuss the new algorithm and the estimated error of the retrieval of tropospheric column ozone; finally, we compare the retrieved tropospheric ozone obtained with the new algorithm with that retrieved by subtracting the stratospheric ozone field from
2.
the archived
TOMS
data.
Current TOMS Algorithm The TOMS
instrument
measures
the radiances
in each of
six wavelengths (312.5,317.5,331.2,339.8,360, and 380 nm) backscattered by the underlying atmosphere and the Earth's surface or clouds [Klenk et al., 1982]. The field of view is about 50 km square. The field of view is scanned, using a mirror, perpendicular to the orbital track in 3ø steps from -51 ø to +51 ø such that contiguous orbits overlap at the
11,137
11,138
HUDSON
ET AL.: DERIVATION
OF TROPOSPHERIC
COLUMN
OZONE
I
i
i
i
Figure 1. Standard total ozone profiles for the low latitudes used in generatingthe total ozone mapping spectrometer (TOMS) lookup tables.
equator. The measured radiances are a function of total column ozone, solar zenith angle, satellite zenith angles (azimuthal angle and scan angle), and the pressure level and reflectivity of the lower boundary. Except for total column ozone and the pressure level and reflectivity of the lower boundary, the other parameters can be derived from the satellite's position. The two longest wavelengths, 360 and 380 nm, where ozone absorption is negligible, are used to derive the reflectivity. The pressureof the lower boundary is obtained either from cloud climatology or from tables of the terrain height versus latitude and longitude averaged over the TOMS footprint. At certain times the TOMS
instrument
measures the solar
flux at the six wavelengthsgiven above, and one can derive the Earth's albedo, A, the ratio of the radiance to the solar
flux. The quantity that is given in the TOMS data archives is the N value, defined as follows:
N = -100.1og10 (A) For a particular choice of solar zenith angle, reflectivity, and for values of the optical depth less than 1.0, N is almost a linear function of the optical depth, simplifying the interpolation routines used in the algorithm. To minimize the effect of calibration errors and of the impact of aerosols or other absorbers, the ratio of the albedos from a pair of wavelengths is used in the algorithm. These pairs were chosen to obtain a large difference in ozone cross section [Klenk et al., 1982]. The algorithm essentially consists of comparing the mea-
sured pair N values with calculated N values for the particular set of solar zenith angle, satellite angles, etc. In practice, this would be a lengthy process and the algorithm consistsof interpolation within a set of precalculatedlookup
tables. For the equatorial region, -30 ø to + 30ø latitude, the lookup tables are calculated for three total ozone amounts of 225, 275, and 325 DU (1 DU is 1 matm cm). The ozone profiles correspondingto these column amounts are shown in Figure 1. The difference between two consecutiveprofiles is mainly in the stratosphere. If we add ozone to the profile, then N will get larger. However, the increase in N is a function of the altitude at which the ozone is placed, and on the reflectivity of the lower boundary. This arises because most of the radiance seenby the instrument comes from the troposphere. If ozone is added to the stratosphereboth the incoming solar flux and the outgoingradiance have passedthrough the added ozone. However, if the ozone is added to the troposphere, this situation does not necessarily hold. The scattered radiance observed by the instrument is the sum of the scattered radiance from the atmosphere and the reflected radiance from the surface. For low-reflectivity surfaces, e.g., the Earth's surface or over the ocean, the contribution of the scattered radiance from the atmosphere dominates and the effective scattering surface for the backscattered radiation to the satellite is in the middle to upper troposphere. If the additional tropospheric ozone is distributed in the lower troposphere, then some of the measured radiance will not have passedthrough the added ozone and the increase in N will be less, i.e., a lower total ozone value will be derived [Klenk et al., 1982]. However, over highly reflecting surfaces, such as clouds or snow/ice cover, the reflected radiance will dominate the scattered radiance, and the efficiency for deriving the additional
total
ozone
amount
will
be close
to that
in the
stratosphere. We say close, because, in general, a component of the reflected radiance from a highly reflecting surface
HUDSON ET AL.' DERIVATION OF TROPOSPHERICCOLUMN OZONE
g
200
scan angle : 0
200
scan
11,139
ang
o U
o
400
400
600
600
o
,_
•
R--0.1
800
.4
R--0.8
• 1000 •
7o../
800
ß./7ø'v ....... I......
1000
0
50
1O0
150
200
0
50
100
Efficiency(%)
0 200
0
scan angle = 50.3
....
200
400
400
600
600 R=
8OO
t=0.8
ß/.P,ø'2'. 0
50
ß
O0
200
i
....
i
....
i
....
scan
800
R 02/
1000
150
Efficiency (%)
.
i
ß
.
,
,
150
Efficiency(%)
1000 0
0••••.•.._,•,.•,•_ 50
100
150
2OO
Efficiency (%)
Figure 2. Retrievalefficiency(sensitivity)of troposphericozonewith respectto reflectivitiesfor four scanangles.Note that ordinaterepresentspressure(altitude)at which 10 DU ozone have been added.R is the reflectanceat the surface(1.0 atm).
in the lower tropospherearisesfrom multiple scattering mbar (International Satellite Cloud Climatology Project) which will tend to increasethe absorptionpath and henceN.
[Rossowand $chiffer, 1991]. Figure 3a gives a plot of the total ozone derived by the current algorithm for a standard assumingthe addition of 10 DU at various altitudes (in profile of 275 DU, versusactual cloud heightsrangingfrom millibars)to the increasein N if the ozonewere placedat the 200 to 800 mbar and for a reflectivity of 0.6. As can be seen ozone peak in the stratosphere.Given that N is linear with from Figure 3a, the archived data will overestimate the total total ozone for the profilesused in the lookup tables, the ozone over the marine stratacumuluscloud by about 15 DU. ratio givesthe efficiencyof the algorithmin detectingtropoFigure 3b shows a similar plot to Figure 3a but for a sphericozone.Thus evenif the troposphericozoneprofileis reflectivity of 0.4. Note that the shape of the curves is the same above a cloud and above the ground, the TOMS different. This is becausethe presentTOMS algorithmalso archived data will show more ozone over the cloud. assumesthat any surfacethat has a measuredreflectivity, R, Figure2 alsoshowsthat the retrievalefficiencyis a sharp between0.2 and 0.6, is the result of partial cloud cover. The functionof the instrumentscanangleat high scanangles. cloud height is then set between the terrain height and the The effect on the archived daily data is to introduce an climatologicalcloud height by the simple relation apparentlongitudinalwave patternwith a period of the orbit P0 = Pcloud --I-(Pterrain-Pcloud)(0.6-- R)/0.4 spacing(---26ø).As the equator crossingof the orbit moves westward each day, this wave pattern will also move westIn fact, although reflectivities over the marine stratocumulus ward. west of Africa vary from 0.4 to 0.7, the visible cloud images As we mentionedearlier, the algorithmdoesnot use actual in this region nearly always show a solid cloud bank. cloud heightsto determinethe pressureof the lower boundIn summary, if the amount of troposphericozone is larger ary but assumesa cloudheightderivedfrom climatology.If than the climatologicalamount used in the TOMS lookup the measuredreflectivityis greaterthan or equalto 0.6, then Figure 2 shows the ratio of the calculated increase in N
thepressure (in atm)of thereflecting cloudsurface,Pcloud, is obtainedfrom the following expression: Pcloud = 0.3 + 0.15 X [ 1 -- cos (2 X latitude)]
At the latitudes of interest in this paper (0ø-20øS)the climatologicalcloud heightis about320 mbar. However, the
tables,the archiveddata will overestimatethe tropospheric contribution to the total ozone above the marine stratocumulus cloud and underestimate it over cloud free land or
ocean. The greater the amount of ozone in the troposphere the greater will be the errors. For southern Africa and the
adjacentAtlantic, troposphericozoneapproaches20% of the total ozone during biomassburning which leads to errors in measured height of the persistent marine stratacumulus the archived total ozone data. The principal effect is to cloud off the west coast of Africa is between 800 and 900 exaggeratethe contrast between the retrieved tropospheric
11,140
HUDSON ET AL.' DERIVATION OF TROPOSPHERICCOLUMN OZONE TOTAL OZONE = 275 D.U., LATITUDE = 15, ACTUAL ALBEDO = 0.6 290
' ' I ' ' ' I ' ' ' a! I 285
280
275
SOLAR ZENITH ANGLE -- 0
SOLAR ZENITH ANGLE -- 30
X : SOLAR ZENITH ANGLE = 60
270
AT SCAN AND AZIMUTHAL
ANGLE=
0
285
400
200
800
600
ACTUAL CLOUDHEIGHT(mb)
TOTAL OZONE = 27.5 D.U., 2go
'
'
I
LATITUDE= 15, '
'
ACTUALALBEDO = 0.4
'
I
'
'
285
280
275
El : SOLAR ZENITH ANGLE = 0
SOLAR ZENITH ANGLE = 30
X : SOLAR ZENITH ANGLE:
270
60
AT SCAN AND AZIMUTHAL ANGLE=
265 200
t
t
t
I
t
,
,
400
I 600
,
,
0
, DO0
ACTUAL CLOUDHEIGHT(mb)
Figure3. Sensitivity of retrieved totalozonein standard TOMSalgorithm to cloudheightfor several zenithangles, for scanangle0ø,at 15ølatitude. Although theretrievalstartswitha standard profileof 275 DU, thefinalmayvaryupto 15DU for a cloudreflectivity of 0.6 (a). Samefor 0.4 cloudreflectivity (b). ozone over the land/ocean and over the cloud. One solution
the extraction of reliable TOMS total ozone highly desirable.
to this ditficultyis not to use data over cloudswith R > 0.4 Oncea methodwas developedfor improvedtotal ozonefor (P. K. Bhartia, personalcommunication,1993). However, thisregionduringthebiomassburningseason,it madesense the tropical SouthAtlantic has becomethe focusof intense to work out an algorithm for TOMS-only tropospheric
experimental activityrecently[Andreaeet al., 1994],making
ozone.
HUDSON
3.
ET AL.: DERIVATION
OF TROPOSPHERIC
Table 2.
New Algorithm Because
of the errors
Residual
in the archived
total
ozone
Since the TOMS
measured
albedos
OZONE
11,141
SAGE and TOMS Data, and Tropospheric Ozone
data
Location
discussedabove, our new algorithm starts with the archived albedos.
COLUMN
are influenced
by both stratosphericand troposphericportions of the total amount of ozone, in order to retrieve one of the components, one must know the other. Both portions are characterized by their column amounts and their profile shapes.In the tropics the retrieval efficiency for stratospheric ozone is close to unity and the archived radiances are largely independent of the shape of the stratospheric ozone profile. However, as shown in Figure 2, the efficiencies and therefore the radiancesdepend on the shapeof the troposphericozone profile. Therefore to retrieve tropospheric column ozone, one must know the stratospheric column ozone and the shape of the tropospheric ozone profile. The new algorithm takes advantage of the relatively smooth variability of stratospheric ozone from the eastern Atlantic, across Africa, to the Indian Ocean. We have confirmed this with SAGE data. The SAGE profiles for September 29 and 30 and October 1, 1989, in this region were integrated down to the tropopause. Table 1 gives a summary of the results. Two features are apparent from Table 1. Stratospheric column ozone has a weak dependence on longitude but a stronger dependence on latitude, and the stratospheric ozone field in this region is relatively smooth with latitude. The SAGE stratospheric ozone was then subtracted from the TOMS total ozone clear sky regions (R < 0.16) to obtain tropospheric ozone (Table 2). In that portion of the Indian Ocean far from the African continent the values obtained for the tropospheric ozone are close to the climatological mean of 28 DU used in the TOMS tables. We have assumed that this climatological mean represents tropospheric column ozone free of the effects of biomass burning. If this is the case, then we need make no corrections to the archived total ozone. Our assumption is that the stratosphericozone field over this region of the Indian Ocean
DATA, DU
Tropospheric Residual
Latitude
Longitude
17øS 18øS 19øS 19øS
3øE 22øE 46øE 70øE
8øS 9øS 9øS
19øE 43øE 67øE
2øN 0øN 0øN
8øW 40øE 64øE
SAGE
TOMS
Ozone, DU
September 29, 1989 262 263 264 267
310 303 296 295
48 40 32 28
September 30, 1989 252 250 252
284 281 284
32 31 32
September 29, 1989 245 246 243
277 269 272
32 23 29
SAGE, StratosphericAerosol and Gas Experiment; TOMS, total ozone mapping spectrometer.
on any day can be obtained by simply subtracting 28 DU from the archived
TOMS
data.
The above technique cannot be used where tropospheric ozone is enhancedby biomassburning, e.g., over Africa and the eastern
Atlantic.
In this case we must use a different
procedure to obtain an estimate of the stratospheric ozone. We first determine
the total ozone column values on either side of the interface between the marine stratocumulus
cloud, 12clou d, and the African continent, •cont, using the correct cloud height. Assuming that both the stratospheric ozone, f•st, and tropospheric ozone, •tr, are the same across that boundary, we can write the following equations for the derived
total ozone:
• cloud =• stq-ecloud • tr
(1)
•cont -- • st q- e cont • tr where e is the efficiency factor derived for the entire tropo-
Table 1.
Derived Stratospheric Ozone for October 1,
1989
Latitude Band, deg South
Stratospheric Ozone Over Indian Ocean, DU
Stratospheric Ozone Over Atlantic Ocean, DU
0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 9-10 10-11 11-12 12-13 13-14 14-15 15-16 16-17 17-18 18-19 19-20
243 243 243 243 245 245 246 249 251 254 256 258 259 260 262 263 265 267 269 271
243 243 243 243 245 245 246 249 251 252 253 255 256 257 259 260 262 264 266 268
spheric profile. We have performed this calculation using measured profile shapes taken from ozone sonde measurements made at Brazzaville (4ø17'S, 15ø15'E) during the 1990 and 1991 biomass burning seasons [Cros et al., 1992]. We can then solve the two equationsfor the two unknowns, f• st and •tr' The values of f•st obtained for October 1, 1989, usingthis techniqueare within 2 DU of thosegiven in Table 2 from the integration of the SAGE profile data. Thus, on any day, we now have two latitude bands of stratosphericcolumn ozone, one at 65øE and the other near 10øE. We then determine the entire stratospheric field from 0øto 65øEby linear interpolation between these two latitude bands (Figure 4). Tropospheric column ozone amounts were obtained using the following procedure: (1) Lookup tables of the expected radiances were calculated for a set of tropospheric and stratospheric column ozone amounts. We used the same stratosphericprofile shapesas in the TOMS algorithm, and the tropospheric ozone profile shape as discussed above (Figure 5). (2) The radiances as a function of latitude and longitude were obtained from the HDT tapes. (3) Using the stratosphericfield obtained as above, the stratosphericcolumn ozone amount was obtained for the particular latitude and longitude by linear interpolation. (4) The tropospheric
11,142
HUDSON ET AL.' DERIVATION
OF TROPOSPHERIC COLUMN
OZONE
EQ 2S
-
4S
244
!
ß
'7
:
.
.
'
6S ß
10S
.•
247
.
25a ,
E66•I .' .......
12S
. ......
. .
,
•
14S
ESg ß
.
15S
.............................
•82
•.-
.
'
ß
18S
..................................
.
...,..
•ae
lOW
õW
0
•[
I0[
I•
.
•
20[
35E
40E
45E
.
50E
551r
Figure 4. Stratosphericozone column•st obtainedfor new algorithmlookup tables. Values are shown at 3-DU intervals. Values for llst at intermediatelongitudesare interpolated.
column ozone amount was then obtained
from the measured
radiances by interpolation in the lookup tables. To avoid problems with the ozone retrieval over clouds, we have limited the reduction of the data to those points for which the cloud height is well known, i.e., above the marine stratocumulus cloud. Over the remainder of the region the clouds tend to be scattered and at random heights and we limited analysisto areas without cloudsin order to provide a consistent picture of total tropospheric ozone. The criteria for this was that the measured reflectivity, as given in the HDT tapes, was less than 0.16.
4.
Error Analysis Three
factors
contribute
to the error
in the retrieval
of
tropospheric ozone. The first comes from the error in the
measured N values, the second from the uncertainty in the stratosphericcolumn ozone, and the third from the uncertainty in the troposphericozone profile. The sources of error in the measured
Thus any variabilityin the derivedozonemapswhich are less than this amount should be treated with caution.
The troposphericcolumn ozone given in Table 2 is within 3 DU of the amount assumedin the TOMS standardprofiles. If our choice of tropospheric column ozone is incorrect, the error produced in the derived stratosphericamount will be only one half of this amount, as the efficiency is about 0.5.
1 oo
i
;
N values has been
discussedby McPeters et al. [ 1993]. The error in N for a pair is constant for the range of values of N used in this analysis and shouldbe of the order of -+0.25. This will give an error in the derived troposphericcolumn ozone of about -+3 DU.
I
i
i
i
I!
' ii
,
, ,
,
,
, , ß
, ,
51
D.U.
lO
Figure 5. Tropospheric ozone profiles used to generatethe lookup tables for the new algorithm [after Cros et al., 1992]. The three ozone profileshave column troposphericozone amountsof 26, 39, and 51 DU.
HUDSON Thus we estimate
ET AL.: DERIVATION
OF TROPOSPHERIC
that the error in our choice of the strato-
spheric column ozone over the Indian Ocean is about -+_ 1.5 DU.
There are only a limited number of ozonesondemeasurements which have been made over tropical Africa. We have compared the retrieved tropospheric ozone using ozone profiles obtained in the 1990 and 1991 biomass burning seasonsat Brazzaville, Congo [Cros et al., 1992], with 1992 Brazzaville and Namibian soundings (Etosha Park at 19ø11'S, 15ø55'E; M. Zunckel and F. Sokolic, personal communication, 1993) taken during the September and October 1992 Southern African Fire-Atmosphere Regional Initiative (SAFARI) and Transport and Atmospheric Chemistry Near the Equator-Atlantic (TRACE A) experiments. Integrated troposphericozone profiles in these casesdiffered by less than 2 DU.
A more detailed analysis of the SAGE data indicates that there are often cases when there is a slight longitudinal gradient in the stratospheric column amount. We have therefore made the assumption that the difference between the amount derived over the Indian Ocean, as described above, and that at the interface is real and represents a longitudinalgradient. This assumptioncould lead to a longitudinal bias to our derived
data.
The method we use to obtain the stratospheric column ozone at the interface
over the West
coast of Africa
intro-
ducesan error of a similar magnitudein this region. Thus we estimate that an error of up to _+6 DU could exist in the gradient of the derived troposphericcolumn ozone between the west coast of Africa
and the Indian
Ocean.
COLUMN
OZONE
11,143
the difference (40 - 28); that is, the new algorithm will derive a value of about (28 + 24) or 52 DU. Both the new and the old algorithms take into account variability in the terrain height in the derivation of the lookup tables. The ozone value is therefore
the column
amount
of
ozone above that terrain height. It is striking that a local ozone minimum appears over a mountainous region of Angola (Plate 1). This may imply a need for further algorithm improvement in this region. It is interesting to note that Thompson et al. [1993] corrected archived total ozone from grid-T TOMS over the marine stratocumulus off southwest Africa. For October 1, 1989, they showed the ozone maximum off the west African coast to be reduced by 10-20 DU depending on reflectivity [see Thompsonet al., 1993, Figure 2]. This is the same as the difference in tropospheric ozone between the new and the old algorithms (compare Plate 1 (top) and Plate 1 (bottom) between 10ø-20øSand 5øW to 15øE). Corrected total ozone over the low cloud obtained by Thompson et al. [1993] appears to be consistent with the new algorithm. However, they did not correct for reduced efficiency in retrieving tropospheric ozone over the continent, thereby retaining an exaggerated ocean-land total ozone discontinuity. The picture that we get from our new analysis is that of an ozone plume extending over much of Africa and the adjacent Atlantic. We are currently analyzing a 2-week record of tropospheric ozone in this region obtained from the new algorithm [Kim and Hudson, 1993]. In particular we are trying to determine dynamic and photochemical influences on areas of high ozone and the persistence of particular features.
5.
Results
and Discussion
Plate 1 (top) showsthe troposphericcolumn ozone derived using our new algorithm for the region from 0ø to 20øS and 10øW to 60øE, on October 1, 1989. The color intervals are 5 DU. Plate 1 (bottom) shows a plot over the same region obtained by subtractingthe stratosphericcolumn field, used in deriving Plate 1 (top), from the archived high-density TOMS (HDT) data. All of the TOMS data was used, there being no restriction on reflectivity as in the new algorithm. This procedure is somewhat analogousto but not the same as the "tropospheric residual" method [Fishman, 1988]. The troposphericresidual is a climatologicaldata product, based on multiday averages of archived TOMS and SAGE data [Fishman et al., 1991]. Our new algorithm is applied only to daily TOMS (HDT) data. Plots of the measured reflectivity and cloud location for this day can be found by Thompsonet al. [ 1993].
There are marked differences between the two pictures. The most obvious differences are (1) the higher ozone over the marine
stratocumulus
cloud as derived
from
the HDT
6.
Conclusions
We have examined the derivation of tropospheric column ozone from TOMS data. We have developed a new algorithm to retrieve tropospheric column ozone directly from the measured albedos for the tropical region where biomass burning introduces large amounts of ozone into the lower troposphere. This method relies on the smoothly varying longitudinaldependenceof stratosphericozone in the tropics and is not likely to be applicable to higher latitudes. We have applied the modified algorithm to the region from 0ø to 20øS and 10øW to 60øE for October 1, 1989, and compared the resultsto troposphericozone derived from the archived total ozone
data.
We
find that methods
based on the archived
TOMS total ozone data underestimate tropospheric ozone by as much as 15 DU over cloud free regions and overestimate it by as much as 20 DU over the persistent marine stratocumulus
cloud off the west coast of Africa.
We have identified three sourcesof significanterror in the new method. There is an error due to the uncertainty in the
ozone data and (2) the smaller contrast between the ozone over the cloud and the continent as derived by the new algorithm. As explained earlier, the first difference is the result of using the incorrect cloud top height. The second
measured albedos, which is estimated to be of the order of +3.0 DU. There is an error in the longitudinal gradient of the ß
results from
DU.
the different
efficiencies
for the derivation
of
troposphericozone by the two algorithms. We would like to emphasize that the difference in efficiency only applies to tropospheric amounts different from the amount assumedin the TOMS standard profiles (28 DU). Thus if a value of 40 DU for the tropospheric column is derived from the HDT data, the efficiency factor discussedearlier only applies to
derived
ozone of less than _+6 DU over 70 ø and another from
the choice of the profile for the tropospheric ozone of +2 Our overall
conclusion
is that the TOMS
radiances
can be
used to derive tropospheric ozone in the tropics when the stratospheric column ozone and the shape of the troposphericozone profile are known. However, archived TOMS total ozone data should be used with caution, especially when daily maps are used to study isolated tropospheric
11,144
HUDSON
Tropospheric
EO
ß
- --0.•-.
.
ET AL.' DERIVATION
ozone ' -.,.....
' --":'
from
the
' . ' '..... ,
ß -.• ..-:5o:-'..'.-':'.-......' : - :•-::..._-:. !,:..
.
:
............
_...•..'%•,.• ..... ''' '••••••••••,>.}.•': '''"' '•-•'"-'"-''" "''''"'"' '' ' '•""• '••••*•:' "•'"'' '•'"'''' ' ''"'' '---'•'' . "'""•'•'•-' '" "••.•'•••••..4.••..' ''"'''"'- -'*''' -
-
-
.-.
'.,
5W
0
Tropospheric
on oct
. ,
-
-
10E
ozone
15E
20E
from
1,
1989
.: .... :"½.,.:':.? ,'"','"'"."'-'.'...'...,,..'...•.....'...,••.....:.•••••,,
. '-'........... "•'•••••••:..:'•' '"' ""•'' ' '"''''' ' '
.........: a•- .4a '•.40 '•"•'-;'."'-" '-,•.• ' ' •'[.-"-v• ......,';-', ' , . ,v.:..'..... _,.•,,:., .....,,•?.?-:,Xo .... ' ., .,'-. ."" -. . ' ...... .•:-*-.-•>-.'•.•t..•'-_..,:.*;A• . . :'..
,:--.--•.... -
5E
...- ."
'"'""'2 '*'""
, 4.",•',, •••.'-'-:,'.". , . ß
- ..,,•-.
, -.%•,•4 ......
-:....: ..::.. ..... -.:....-..•. ............. _..........'.,: '' ",.--'.-'.'.. '. 35 . ,.---•:.....,....••12:•..•. ß•'>'2•_. •..':; ,_•,• •.;•:
lOW
OZONE
'•':'"':::> •:•" •":,,"-"•-' ',,."...-'-d:..•"-:4:,..:#:..•...•r 'i':'. •:.•..•. •
'•::'*"• '. '-•:? "••••i:;.t:•-•---"•--'-•-••,...,.... '•' ........... '""'"" *'''"' "'*• '"'• '-• '"""" ••••...,•.,,.•> ß-
COLUMN
new algorithm :
•:::-:-"'"-'-"-••/•,':':*'•.:•. "'- ...... :;.,,': ..... ';':'-.'-:-.•,40'•. :" ½.......... :
']"" •"•_ "-.... -*,•- '-•'-- ß •. 14S'•,• "•'•'•'•, '••••'1 '"'""'•'• ' -"-'•. '••
OF TROPOSPHERIC
:•-:"-'::• ........ ;..:'"'"'"'>':'•:4'•"" ...."?'"':•':• b""¾'-" '•' '•••••••:.,. "''' '":' :.'ß-"-ßß-'ßß ....... .'-•:;?;: ß.. ..`....•...:.."``.........`...........•....•.....••
.. -* .. ..-.. -.....
-:.....>.::•'.?• :½.... .... 4-.-..-.....
.::.....v...-0•.. . =:,-'"":",.'•:•:::1;: "'.":":•., '?..,:.,.,.,: - '-'. ,..... .. :.-• ,?.:t•' .
25E
30E
erchived
35E
40E
45E
TOMS on oct
4õE
50[
1,
50E
55E
1989
55E
boUo
Plate 1. Tropospheric ozone column over Africa, western Indian Ocean, and eastern Atlantic for October 1, 1989. Each color bar representsan increment of 5 DU. (top) Retrieved by the new algorithm. (bottom) Obtained by subtracting stratospheric ozone derived from the Stratospheric Aerosol and Gas Experiment instrument from the archived total ozone (old algorithm). Major differences in the top panel compared to the bottom panel are the reduction in ozone over the marine stratocumuluscloud off the West African coast (up to 20 DU) and an increase over the African continent (frequently 5-10 DU).
ozone features.
Errors
in ozone from
the archived
data are
substantial (to 15-20 DU) when the lower troposphere has 2-3 times greater ozone than in the standardTOMS profiles or over a low, highly reflecting surface like marine stratocumulus. Errors are also expected to be significantat extratropical latitudes. Acknowledgments. This work was supported by the NASA Atmospheric Chemistry Modeling and Analysis Program grant NAGW-1696 to the University of Maryland and through funding to Goddard Space Flight Center. Our thanks are due to Jack Fishman, who inspired this study, and to P. K. Bhartia for helpful comments on the original manuscript.
References Andreae, M. O., et al., Biomass burning in the global environment: First results from the IGAC/BIBEX field campaign STARE/ TRACE-A/SAFARI-92, in Global Atmospheric-biosphericChemistry: The First IGAC Scientific Conference, edited by R. Prinn, Plenum, pp. 83-401, New York, 1994. Cros, B., D. Nganga, A. Minga, J. Fishman, and V. Brackett, Distribution of tropospheric ozone at Brazzaville, Congo, determined from ozonesonde measurements, J. Geophys. Res., 97, 12,869-12,876, 1992.
Dave, J. V., and C. L. Mateer, A preliminary study on the possibility of estimating total atmospheric ozone from satellite measurements, J. Atmos. Sci., 24, 414-427, 1967. Fakhruzzaman, K., J. Fishman, V. Brackett, J. D. Kendall, and
HUDSON
ET AL.: DERIVATION
OF TROPOSPHERIC
C. O. Justice, Large-scale circulation patterns associated with high ozone concentrationsof tropospheric ozone in the tropical south Atlantic Ocean, in Ozone in the Troposphere and Stratosphere, edited by R. D. Hudson, NASA Conf. Publ., 3266, 158-161, 1994.
Fishman, J., Tropospheric ozone from satellite total ozone measurements, in Tropospheric Ozone, edited by I. S. A. Isaksen, pp. 111-123, D. Reidel, Norwell, Mass., 1988. Fishman, J., and J. C. Larsen, Distribution of total ozone and stratosphericozone in the tropics: Implications for the distribution of tropospheric ozone, J. Geophys. Res., 91, 6627-6634, 1987.
Fishman, J., P. Minnis, and H. G. Reichle Jr., The use of satellite
data to studytroposphericozone in the tropics,J. Geophys.Res., 91, 14,451-14,465, 1986.
Fishman, J., F. M. Vukovich, D. R. Cahoon, and M. C. Shipham, The characterisation of an air pollution episode using satellite total ozone measurements, J. Clim. Appl. Meteorol., 26, 16381654, 1987.
Fishman, J., C. E. Watson, J. C. Larsen, and J. A. Logan, The distributionof troposphericozone determinedfrom satellitedata, J. Geophys. Res., 95, 3599-3617, 1990. Fishman, J., K. Fakhruzzaman, B. Cros, and D. Nganga, Identification of widespread pollution in the Southern Hemisphere deduced from satellite analyses, Science, 252, 1693-1696, 1991. Fishman, J., V. G. Brackett, and K. Fakhruzzaman, Distribution of tropospheric ozone in the tropics from satellite and ozonesonde measurements, J. Atmos. Terr. Phys., 54, 589-597, 1992. Hudson, R. D., and J. Kim, Direct measurementsof tropospheric 03 using TOMS data, in Ozone in the Troposphere and Stratosphere, edited by R. D. Hudson, NASA Conf. Publ., 3266, 119-121, 1994.
Kim, J., and R. D. Hudson, Measurementsof tropospheric ozone during the biomass burning seasonin Africa and South America using TOMS radiances and SAGE data, Eos Trans. AGU, 74(16), Spring Meet. Suppl., 104, 1993. Klenk, K. F., P. K. Bhartia, A. J. Fleig, V. G. Kaveeshwar, R. D.
COLUMN
McPeters, and P.M.
OZONE
11,145
Smith, Total ozone determination from the
BackscatteredUltraviolet (BUV) experiment, J. Appl. Meteorol. 21, 1672-1684, 1982.
Logan, J. A., and V. W. J. H. Kirchhoff, Seasonal variations of troposphericozone at Natal, Brazil, J. Geophys. Res., 91, 78757881, 1986.
McPeters, R. D., et al., Nimbus-7 Total Ozone Mapping Spectrometer (TOMS) data products user's guide, NASA Ref. 22 Publ. 1323, Washington, D.C., 1993. Pickering, K. E., A.M. Thompson, D. P. McNamara, M. R. Schoebed, P. A. Newman, L. R. Lait, C. O. Justice, and J. D. Kendall, A trajectory modeling investigation of the biomass burning-tropical ozone relationship, in Ozone in the Troposphere and Stratosphere, edited by R. D. Hudson, NASA Conf. Publ., 3266, 101-104, 1994a.
Pickering, K. E., A.M. Thompson, D. P. McNamara, and M. R. Schoebed, An intercomparison of isentropic trajectories over the South Atlantic, Mon. Weather Rev., 122, 864-879, 1994b. Rossow, W. B., and R. A. Schiffer, ISCCP cloud data products, Bull. Am. Meteorol. Soc., 72, 2-20, 1991. Thompson, A.M., K. E. Pickering, D. P. McNamara, and R. D. McPeters, Effect of marine stratocumulus on TOMS Geophys. Res., 98, 23,051-23,058, 1993.
ozone, J.
Watson, C. E., J. Fishman, and H. Reichle Jr., The significanceof biomass burning as a source of carbon monoxide and ozone in the southern hemisphere tropics: A satellite analysis, J. Geophys. Res., 95, 16,443-16,450,
1990.
R. D. Hudson and J.-H. Kim, Department of Meteorology, University of Maryland, 2213 Computer and Space SciencesBuilding, College Park, MD 20742. A.M. Thompson, NASA Goddard Space Flight Center, Code 916, Greenbelt, MD 20771.
(Received March 21, 1994; revised September 16, 1994; accepted September 16, 1994.)
