The original concept of deriving tropospheric column ozone in the tropics by a residual method consists of using total column ozone from TOMS (the Total Ozone Mapping Spectrometer) and subtracting stratospheric column ozone from an independent sensor. It is assumed that stratospheric ozone is invariant or nearly constant with longitude so that variations in total ozone detected by TOMS can be attributed to variations in tropospheric column ozone.

     There are three general limitations of this approach. First, there may be a mismatch in orbital and sampling characteristics between TOMS and the other sensor. This occurs with SAGE (Stratospheric Aerosol and Gases Experiment), which typically samples the 10N-10S latitude band 50-60 days per year, sometimes missing the tropics for several months in a row. Thus, the original tropospheric ozone residual technique of Fishman and coworkers [Fishman et al., 1990; 1991; update in Fishman and Brackett, 1997], which used SAGE, reported climatology based on 2-3 month-averaged tropospheric column ozone. Second, TOMS total ozone does not detect near-surface ozone over surfaces with low albedo with 100% efficiency [Hudson et al., 1995; Kim et al., 1996; Hudson et al., 1998]. Third, stratospheric column ozone derived from the independent sensor may be uncertain in the lower stratosphere. This is a particular concern for SBUV (Solar-Back- Scatter Ultraviolet; Ziemke and Chandra, 1998; Hudson et al., 1998) and to a lesser degree for the UARS (Upper Atmosphere Research Satellite) MLS (Microwave Limb Sounder) and HALOE (Halogen Occultation Experiment). The two latter sensors have been used with TOMS in a residual approach [Ziemke et al., 1998].

     These factors have motivated the development of two methods that use TOMS total ozone with physical parameters other than a separate satellite measurement to distinguish stratospheric ozone from tropospheric ozone. The two methods, the convective- cloud-differential (CCD) method [Ziemke et al., 1998] and the modified-residual approach [Hudson and Thompson, 1998], are described in detail in the 20 Sept 1998 Journal of Geophysical Research.

     In the CCD method, stratospheric ozone is set equal to the TOMS total ozone measurement over highly reflecting, high altitude clouds in the western Pacific Ocean. At cloud-free pixels, signified by reflectivity < 0.2, tropospheric column ozone is obtained by subtracting the above-cloud stratospheric ozone amount from TOMS total ozone. This is referred to as the CCD tropospheric ozone column. Because there is no cloud height information in TOMS, a primary assumption in the CCD method is that the high reflectivity clouds have cloud tops at the tropopause. Monthly averaged maps between 20N and 20S are shown in Ziemke et al. [1998]; no correction is made for reduced detection efficiency near the surface.

     In the modified-residual (MR) method, tropospheric column ozone taken from ozonesondes near the Atlantic ozone maximum is subtracted from TOMS total ozone to give the stratospheric column ozone value [see Appendix]. Stratospheric column ozone subtracted from total ozone yields tropospheric ozone at all other longitudes; an efficiency correction is applied because only cloud-free (low reflectivity) TOMS data are used. Fourier analysis of an underlying wave-one pattern in total ozone [Kim et al., 1996] sets the latitudinal limits of a tropical air mass within which the MR method is assumed to be valid (Figure 4 in [Hudson and Thompson, 1998]).

     The MR method was based on a series of ozonesondes in the vicinity of the ozone maximum which were only available for 1991- 1992, during the pre-TRACE-A (Transport and Atmospheric Chemistry near the Equator - Atlantic) and SAFARI (Southern African Fire Atmospheric Regional Initiative)/TRACE-A intensive. A seasonal regularity observed in the 1991-1992 ozonesonde data (Figure 5 in Hudson and Thompson, 1998) suggested that the method might be applicable to other years. We tried the method for 1990 and the comparison between derived tropospheric ozone and sondes looked promising. From that point, the modified-residual method was extended back in time to cover the entire Nimbus 7 (=N7)/TOMS record (1979-1992, Appendix), assuming that the 1991-1992 background ozone climatology was applicable. For this purpose, TOMS total ozone was averaged over two 13- to 16-day periods per month. These images, referred to as N7/TTO maps, appear on this website (http://metosrv2.umd.edu/~tropo). An example appears in Plate 1A.

     In early 1997, we began using the MR method to process real- time Earth Probe (EP) and ADEOS (Advanced Earth Observing System) data. Daily, 3-day- and 9-day-averaged images also appear on the homepage: http://metosrv2.umd.edu/~tropo/(DATA). A sample ADEOS map is given in Plate 1B. It compares well with the only near-tropical ozonesonde data available for the period shown.

     This paper presents the first evaluations of the pre-1991 N7 and post-1996 EP/TTO maps through comparisons with ozonesondes. The study focuses on the Atlantic and adjacent continents (Section 2) because this region has the greatest amount of near- equatorial tropospheric ozone data for the period prior to 1991 and after 1996: Natal (Brazil, 1979-1990), Ascension Island (1990, 1997-1998), Brazzaville (1990).

     Potential applications of TTO maps are illustrated by analyses with Atlantic and near-Atlantic regional time-series which are used to address the following questions (Section 3):



(1) What do seasonal cycles and interannual variability in TTO look like and how do they compare with those observed in ozonesondes?
(2) Do TTO and derived stratospheric ozone, O₃)_str, show signals typical of ENSO influences?
(3) Can trends in tropospheric ozone over eastern South America and the eastern Pacific be detected for the 1980's, as reported using indirect approaches with TOMS ozone [Jiang and Yung, 1996; Kim and Newchurch, 1996]?


Answers to these questions demonstrate that the MR method gives TTO maps that may be useful for tropical climatological studies and time-series analyses. Furthermore, with 1x2 degree resolution and 2-3 maps/month, TTO maps by the modified-residual method are more highly resolved spatially and temporally than tropical tropospheric ozone obtained by other approaches to satellite data. In particular, they may be useful for process studies and field campaigns.

A. Stratospheric Ozone

     Equatorial stratospheric column ozone, O3)str, derived from the MR method agrees, to within 8 DU, with pre-Mt Pinatubo- eruption SAGE II measurements recorded between 10N and 10S in 1985-1991 (Figure 1). Slightly less good agreement is found for derived stratospheric ozone and ozone integrated from 1-100 hPa from the UARS/MLS launched in September 1991. Discrepancies between the N7 derived stratospheric ozone from September 1991- December 1992 compared to MLS stratospheric O3 are similar to those from the EP period shown in Figure 2. The fact that O3)str is lower than the SAGE and MLS stratospheric column values may reflect systematic differences in effective tropopause height. Integrated tropospheric ozone from the sondes at Ascension, Brazzaville and Natal includes balloon data from the surface to 100-120 hPa. This is where the ozone gradient changes sharply (see, for example, Figure 5 in Thompson et al., 1996a) and it also corresponds to the tropopause typically reported in NCEP analyses. If the SAGE or MLS tropopause is effectively 3 km lower than the NCEP tropopause, it would be 6-10 DU greater than integrated ozone from the sounding. Note, however, that differences of 8-15 DU between O3)str and observations from the other satellites are within the precision of SAGE and MLS column ozone below 20 km [WMO, 1998].

     There may be a systematically high bias to MLS stratospheric ozone. Figure 3 (*) shows stratospheric ozone based on the Ascension sondes. Integrated tropospheric ozone is subtracted from TOMS total ozone to obtain stratospheric column ozone. (Stratospheric column ozone from the sondes is not used directly because of potential inaccuracies from extrapolating ozone above sonde burst altitude). For the months for which both are available (July 1997-April 1998), MLS ozone is 5-18 DU (Figure 2, July-October 1997) greater than stratospheric ozone based on the sondes. Accordingly, agreement between derived stratospheric ozone, O3)str, and stratospheric ozone from the sondes is better (mean deviation=6 DU in Figure 3) than with MLS. Deviations from sonde-based stratospheric ozone are more evenly distributed between positive and negative values. For the Nimbus 7 period, derived stratospheric ozone, O3)str, compares well with stratospheric ozone based on Natal ozonesondes (RMS deviation = 8.8 DU for 1979-1992; Figure 4). Here too, a comparison of stratospheric ozone from the sondes and SAGE stratospheric ozone (Figure 1) shows the latter to be higher than stratospheric ozone inferred from the sondes; however, the offset is only ~5 DU.

     The CCD method also derives stratospheric column ozone appropriate for the tropics [Ziemke et al., 1998]. A comparison of derived stratospheric ozone from the MR and CCD methods and averaged within 10 degrees of the equator, shows that agreement between the two techniques is within 4 DU, which is remarkably good, considering how different the assumptions are when deriving the stratospheric ozone column. A signature of the QBO (Quasi- Biennial Oscillation) appears in both records.

B. Tropospheric Ozone (TTO)

     Evaluation of tropical tropospheric column ozone during the pre-1991 Nimbus 7 period and during ADEOS and EP is complicated by a geographically and temporally uneven ozonesonde data base. For the 1980's, only Natal ozonesonde data are readily available [Kirchhoff et al., 1991, 1996; Logan, 1994]. For comparison with the sondes, TTO from the MR method is averaged over the 9 pixels from 5-7S and 32-38W. The mean agreement with the TTO obtained at Natal (mean deviation = 7.9 DU, Figure 5) is good, considering that the sonde record of 1-3 sondes/month (except during intensive campaigns [Kirchhoff et al., 1996]) may not capture accurate half-month averages. The same level of agreement with Natal sondes was found with the CCD method (Figure 7b in Ziemke et al., 1998).

     Figure A-2A shows satisfactory agreement between twice monthly averaged TTO from Nimbus 7/TOMS (1990 and 1991) and integrated tropospheric ozone from sondes launched at Ascension Island before the TRACE-A field experiment. The RMS agreement between the sondes and TTO averages 5.1 for 1990 and 5.3 for 1991. The 1991 comparison (also in Hudson and Thompson, 1998) includes the referencing of background tropospheric ozone to the Ascension data (as in Figure A-1). However, the 1990 comparison is made with independent data, as is the 1997-98 comparison shown in Figure 6. Figure 6 uses twice-weekly Ascension sondes launched from mid-1997 through mid-1998 and 9-day-averaged TTO. Agreement in Figure 6 is similar to that for 1990 and 1991: 5.1 DU mean deviation between the sondes and the EP/TTO record.

     In Figure 7, the comparison between TTO and ozonesonde data from American Samoa (14S, 171W; Komhyr et al., 1989) is performed by extending the MR method beyond strictly tropical air, as indicated by Fourier analysis of the wave-one pattern. Accordingly, agreement between TTO and Samoan ozonesondes during 1986-1990 is not as good as at Natal or Ascension; the mean deviation is 11.8 DU, with TTO usually lower than Samoan tropospheric ozone. Using the 1995-1996 PEM-Tropics ozonesondes, Folkins et al. [1999] have shown that the upper troposphere over Samoa between 14 and 17 km includes air with a considerable stratospheric signature. By this criterion, Samoa would fall outside the band of strictly tropical air and the modified- residual method would not produce the most accurate results. Stratospheric air mixed into the upper troposphere would be expected to give higher tropospheric column ozone than TTO, which assumes a distinct stratosphere and troposphere. This is the usual direction of disagreement between the two.

3. Evaluation of Nimbus 7 TTO through Regional Time-Series Analysis

     In this section, the quality of TTO and other parameters derived in the MR method is evaluated by using time-series to address questions about climatology and atmospheric processes that affect tropical ozone. Most of the time-series are based on N7 data because the EP/ADEOS data records are less than three years. Furthermore, the dominance of the 1997-1998 ENSO in the EP tropospheric ozone record [Chandra et al., 1998] probably renders the EP time-series atypical. Nimbus 7 TTO data during the period in which total ozone was suppressed by the Mt Pinatubo eruption (mid-1991 through June 1993) are also used sparingly.

A. Seasonal Cycles and Interannual Variability at Ozonesonde Sampling Sites

     Time-series of TTO data over given regions are used to address the question:
What do seasonal cycles and interannual variations in TTO look like and how do they compare with variations observed in ozonesondes?

     Figure A-2A shows comparisons between ozonesonde data in the Atlantic region and TTO for the 9 pixels surrounding Ascension, Brazzaville and Natal (also Figure 5 for the latter). The distinctive feature at all three sites is a regular seasonal variation, with tropospheric ozone peaking 20-25 DU higher in the latter half of the year than in the second quarter when ozone is a minimum.

     Figures 8A, 8B, 8C show time series of TTO at the Natal, Ascension and Brazzaville locations, along with larger regions between the eastern Pacific just west of South America across the Atlantic to southern Africa (Figures 8D, 8E, 8F, 8G). The seasonality of TTO in the regions shown in Figure 8 is determined by fitting a linear regression model [Hollandsworth et al., 1995] to the 14-year twice-per-month time-series. The model, which has been used extensively for analysis of trends in total ozone [Stolarski et al., 1991], permits the assumption of a seasonal cycle and linear trend (Section 3B). The deseasonalized mean, assumed to pertain to the beginning of the 14-year record, is shown in each panel of Figure 8. The regression model can be modified to take into account solar cycle effects, ENSO signals and the QBO; this has not been done in the current analysis. The reader is referred to Ziemke et al. [1998] and Thompson et al. [1996b], respectively, for discussion of QBO effects on derived stratospheric ozone and total ozone in the Nimbus 7 records.

     Figure 9A,9B,9C shows the estimated seasonal cycle from the regression model best-fit for TTO (solid line) and ozonesonde data (----) at the sounding sites. For Natal (Figure 9A), there are 155 of a maximum possible 336 data points based on twice-per- month averaging of the sondes over 14 years (see * in Figure 5). The coherence of the TTO and sonde seasonality for Natal is very good. There is agreement among all points in the time-series within the stated error limits (± 5 DU) for both TTO and the averaged sondes. The highest tropospheric ozone occurs in the second half of the year. This corresponds to the season of savanna burning, which is believed to account for much of the excess ozone [Fishman et al., 1991].

     For Ascension and Brazzaville, there are only 39 and 42 data points compared to 72 twice-monthly intervals in the 1990-1992 record. The Ascension ozonesonde record is too sparse in the first half of the year to use in the linear-regression analysis and the amount of Brazzaville data is marginal. If one consults Figure 8b in Thompson et al. [1996b], it is seen that 1990-1992 Ascension data, averaged for each month, show tropospheric column ozone with two peaks in the June to September period. Ascension TTO is a broad peak between August and October. For Brazzaville the 1990-1992 ozonesonde data [Thompson et al., 1996b] show a broad peak between July and late September, with a suggestion of two peaks. Two peaks are more distinct in the seasonal TTO, one corresponding to the strongest period of burning in south central Africa (June-July) and the second a few weeks after the September peak seen in the sondes.

     In summary, more sonde data are required for evaluating TTO seasonality than are available at present. Only the Natal data cover enough years for a statistically meaningful comparison. Agreement appears to be good, but even at this location, interpreting tropospheric ozone seasonality is not straightforward. During SAFARI/TRACE-A, excess tropospheric ozone over Natal, with excess defined as any ozone greater than a non-burning season column ozone value of 31.5 DU, appeared to originate from Africa 60-70% of the time [Thompson et al., 1996a]. Towards the dry to wet season transition (October- November), tropospheric ozone from other NO sources besides biomass burning, e.g. lightning and soil releases, might also enhance ozone. Dynamical patterns can redistribute free tropospheric ozone over days to weeks [Krishnamurti et al., 1996]. These factors may cause tropospheric ozone at Natal to peak 1-3 months later than the biomass burning which is fueling much of the ozone. The appearance of two peaks in the Ascension and Brazzaville TTO could also reflect interaction of multiple processes affecting ozone.

     Time-series over the Nimbus 7 period are used to explore the question:
Do the wave amplitude and TTO derived from the MR method show signals expected from ENSO influences?

     The amplitude of the wave-one feature (Figure A-1) from 1979-1992 is a measure of the contrast between the Atlantic background tropospheric ozone maximum, O₃)_tr^back(0), and the Pacific minimum, O₃)_tr^back(180). In Figure 10, the open circles show the wave magnitude normalized to the mean. The maximum wave amplitude in the March-May period (cf Figure 7 in Hudson and Thompson [1998]) corresponds to greater convective activity over the Pacific Ocean, which brings ozone-poor air from surface up through the free troposphere, reducing the tropospheric ozone column amount. In this respect, the wave may be thought of as signifying an eastern Pacific ozone deficit relative to the Atlantic. In the August-October period the maximum wave amplitude reflects "excess" ozone over the Atlantic, which is attributed to advection of photochemically produced ozone from the adjacent continents, followed by subsidence [Chatfield et al., 1996; Krishnamurti et al., 1996; Thompson et al., 1996a]. The minimum wave amplitude occurs in approximately half the Junes and Decembers in the 14-year Nimbus 7 record. Although Atlantic TTO remains higher than Pacific TTO in most of these cases, the underlying background tropospheric ozone difference is believed to disappear during the semi-annual migration of the ITCZ across the equator. These correspond to the minimum values in Figure 10, which have been normalized to the mean; the wave amplitude is zero at these minima.

     Correlations between interannual variations in the total ozone wave-one [Shiotani, 1992] and in tropospheric ozone in the Nimbus 7 period have been discussed elsewhere and are not repeated here [Kim and Newchurch, 1996; Ziemke et al., 1998]. The positive correlation of wave amplitude with one indicator of ENSO anomalies (in sea surface temperature, SST) in Figure 10 is similar to that reported in other studies. The ENSO episodes where this is seen the best are in the early part of the 1982- 1983 ENSO and the latter part of the 1987-88 ENSO. The SST anomaly shown is based on statistics from one of several Pacific regions. It suggests that the MR method is consistent with other approaches of isolating the wave-one pattern and of distinguishing stratospheric and tropospheric ozone. Dynamical effects on upper tropospheric ozone during the intense 1997-1998 ENSO event have been described by Chandra et al. [1998].

     Variability of TTO with respect to the 1982-83 ENSO depends on the region of focus. Figures 5 and 8A suggest little impact at Natal. Unfortunately, this is hard to verify with the sonde data because late 1982 to early 1983, the period of greatest ENSO disturbance, coincides with a paucity of Natal ozonesondes. Figures 8E and 8F, corresponding to eastern South America (mostly Brazil, 0-12S, 40-70W) and the south Atlantic (0-12S, 0-40W), respectively, span larger regions east and west of Natal. For the south Atlantic, which includes Natal, the 1982-83 ENSO does not stand out as having high ozone. Natal and the south Atlantic show very low TTO toward the end of the 1982-83 ENSO, however. The 1982-83 ENSO corresponds to the highest TTO during the Nimbus 7 period in the eastern South America location (Figure 8E). Southern African TTO (Brazzaville, Figure 8B; 0-12S, 0-30E; Figure 8G) is the opposite. TTO was a 14-yr low during the period of the usual seasonal ozone maximum; the area averaged TTO never exceeded 45 DU during the 1982 dry season when TTO over eastern South America was > 70 DU.

     The eastern Pacific was exceptionally low in TTO during the 1982-83 ENSO (Figure 8D). The reasons for the latter are probably a combination of dynamics and photochemistry. Winds that transport ozone and ozone precursors westward from eastern South America, may have been modified in the 1982-83 period [Kim and Newchurch, 1996]. In addition, when convection brings additional H₂O into the upper troposphere along with low ozone, photochemical ozone destruction takes place if NO, naturally low at the surface in the eastern Pacific [Thompson et al., 1993], remains low. Suppression of tropospheric ozone under the influence of convection has been observed in sonde profiles taken over the Pacific in early 1993 [Kley et al., 1996].

     Evidence for the ENSO in 1997-1998 may be apparent in TTO at Ascension Island, as well as from the ozonesondes launched there. The mean TTO at Ascension for the 1979-1992 record (Figure 8B) is 38.4 ± 0.62 DU whereas during the peak of the 1997 ENSO, mean TTO was 40 DU and integrated tropospheric ozone from the sondes averaged 43 DU. Further evidence for the ENSO influence in TTO comes from a comparison of 1979-1992 Nimbus 7/TTO with means from TTO obtained from EP/TOMS between July 1997 and February 1998, the months of most pronounced ENSO effect. Zonally averaged, TTO between 0 and 12S is about 20% greater in the ENSO 1997-1998 period than during 1979-1992. However, over eastern South America and at Natal, the 1997-1998 ENSO period was 8-10 DU or up to 30% higher than 1979-1992. Southern Africa was not remarkably different in late 1997 and early 1998. The most dramatic effects in TTO during the 1997-1998 ENSO, verified by ozonesonde observations [Fujiwara et al., 1999], were over southeast Asia, due to the biomass burning over Indonesia [Thompson et al., 1998]. Means for various regions for 1979-1992, compared to the most intense part of the 1997-1998 ENSO, are as follows:

     Comparison of EP/TOMS with the ground-based Dobson network of instruments for measuring total ozone has shown that EP/TOMS is more consistently higher relative to the Dobson network than was N7/TOMS (G. Labow, R. McPeters and R. Stolarski, personal communication, 1998). This would imply that 4-5 DU higher TTO in EP/TOMS relative to Nimbus 7 could derive from an instrument offset. Thus, we do not make detailed comparisons of the N7 and EP records at this point, except to note that the eastern South America, Natal and Indonesian ENSO signals for 1997-1998, relative to the N7 record, exceed any possible instrument artifact.

     In contrast to residual methods [Fishman et al., 1991, 1996; Ziemke et al., 1998] which are based on multiple sensors, with mismatches in sampling period, footprint, calibration and operational lifetime, tropospheric ozone determined from the TOMS-only MR method is well-suited for analysis of trends.

     Because of the brevity of the EP/TOMS TTO and its strong ENSO signature, only TTO data from the N7 record are used in trends analysis here. We also note that from mid-1991 through 1992, record low total ozone, due to the eruption of Mt Pinatubo, may introduce a strong bias in the latter part of the N7 record. For this reason, and because the first year of Nimbus 7 operation appears to have anomalously low tropospheric ozone, only TTO from 1980 through 1990 are used for analysis of trends. The linear regression model [Hollandsworth et al., 1995] used for determination of seasonality (Figure 9A,B,C) is the basis for determinating the linear and seasonal trends with the N7 TTO record. Model results for Natal appear in Figure 11A,B. The model fit (----) to the TTO time-series (solid line in Figure 11A) is excellent. The ozonesonde data with model fit appear in Figure 11B. The straight line represents the linear trend, with the deseasonalized mean TTO and sonde integrated tropospheric ozone values in Table 1, as the starting points in Figures 11A,B. There is no significant linear trend at Natal, according to the TTO record: -0.09 ± 0.16 DU/yr; the Natal mean TTO over 1980-1990 is 38.8 ± 1.1 DU. The sonde data, which are quite sparse (see also Figure 5), give a mean 6 DU less than TTO (Table 1) and no significant trend: 0.33 ± 0.42 DU/yr. This can be compared to the analysis of Logan [1994], who finds a slight increase in the Natal ozone vertical profile, which is only statistically significant between 600 and 400 hPa [UNEP, 1998].

Table 1. Regional TTO (DU) and Trends (DU/yr) in Tropical
Tropospheric Ozone from MR method, 1980-1990

	Mean column TTO (DU, ± 2)	Trend (DU/yr) (± 2)	Figure No.
Natal, sondes   6S, 35W	33.4(2.8)	+0.33(0.42)	11B
Natal, TTO   5-7S, 32-38W	38.8(1.1)	-0.094(0.16)	11A
Ascension Is. TTO   7-9S, 12-18W	39.6(1.2)	-0.015 (0.17)	12A
0-12N, zonal mean	29.9(.68)	-.02(.10)	n/a
0-12S, zonal mean	30.3(.58)	-.054(.08)	n/a
East S. America   0-12S, 40-70W	37.1(1.1)	-0.11(.16)	12B
E. Pacific   0-127S, 80-110W	27.9(.88)	-0.11(.16)	12D
S. Africa   0-12S, 0-30E	40.3(.92)	-.02(.14)	n/a
S, Atlantic   0-12S, 0-40W	40.4(.84)	-.06(.12)	12C

     The linear regression analysis of TTO for other regions shows that seasonality, interannual variability and trends at Natal are typical of tropospheric ozone over the entire south Atlantic Basin. Figure 12A,B,C,D, shows TTO corresponding to Ascension Island, eastern South America, the South Atlantic, and the Eastern Pacific, with the model fit for each case and the linear trend. There are no significant trends on an annual basis (Table 1), but small seasonal trends (not shown) may be marginally significant in the first part of the year. The South Atlantic region encompasses both Natal and Ascension Island, whereas South America represents most of northern Brazil. For completeness, analysis of the 11-year TTO record in the zonally averaged band from 0-12S and 0-12N is also given in Table 1. No significant trend is apparent throughout the tropical band.

     The regression analysis suggests that tropospheric column ozone in regions well-known for seasonal burning - Brazil, and south central Africa, for example - did not change significantly during the 1980's. Is this consistent with correlative changes in biomass burning? A number of satellite sensors for detection of fires were operational in the 1980's, but no single fire count product is available to compare with the N7 TTO record. Although smoke as a proxy for fires is difficult to correlate consistently with tropospheric ozone due to different characteristics of low- altitude detection of ozone and aerosols, the TOMS absorbing aerosol product (which detects smoke and dust) is available for the entire N7 period (see, the TOMS homepage: http://toms.gsfc.nasa.gov to view monthly-averaged absorbing aerosol maps). Hsu et al. [1999] infer a possible 1980's increase in smoke aerosol at several South American and African locations at which sun photometer data have been used to calibrate the TOMS aerosol product. Accordingly, we might expect TTO to increase in regions for which biomass burning is a major ozone source, e.g. over the Atlantic and adjacent continents. On the other hand, CO, a major product of biomass burning, has not shown an increase at tropical flask sampling sites, although observations began at most of these in the late 1980's and early 1990's [Novelli et al., 1998]. We are currently looking at the 1980's absorbing aerosol index from TOMS for the regions in Table 1 to see whether any trends emerge and to compare the seasonality of smoke aerosols and TTO.

     The finding of no significant trends in tropical tropospheric ozone agrees with Chandra et al. [1999], who have examined the 1979-1992 tropical tropospheric ozone time-series obtained by the CCD technique. Note, however, that in either approach the magnitude of TTO trends is at the edge of the stated precision limits. Over a decade, a 1%/year increase in tropospheric column ozone would be 2-3 DU. How do our results compare to trends in lower tropospheric ozone (surface to 6 km) inferred from TOMS by a terrain-differencing method [Jiang and Yung, 1996; Kim and Newchurch, 1996; 1998]? These groups have analyzed smaller regions from 1979-1992:

     Our inference of no trend for the eastern Pacific (Table 1) between the equator and 12S concurs with Kim and Newchurch [1996]. However, the TTO maps show no increases east of the Andes during the 1980's (eastern South America, Figure 12B), whereas Kim and Newchurch [1998] find a 1%/yr increase. There are several possible explanations for this. Kim and Newchurch [1996, 1998] use only a few pixels of TOMS data for their time series. They also used the gridded TOMS product, which is subject to artifacts due to clouds and low-altitude ozone detection efficiency [Hudson et al., 1995; J. R. Ziemke, P. K. Bhartia and R. D. Hudson, unpublished results, 1998].

     A time-series of maps based on TOMS ozone data, two per month with a 1° latitude by 2° longitude grid, have been produced for tropical tropospheric column ozone between 15N and 15S. Periods include the 1979-1992 Nimbus 7 record and ADEOS and EP-TOMS, the latter from 1996 to the present. The technique used, the modified-residual method, is based on high density, cloud-free TOMS ozone in combination with a climatology of 1991- 1992 ozonesonde data to separate stratospheric and tropospheric ozone. In this study, the focus is on the tropical Atlantic and near-Atlantic stations. The major findings are:

(1) Evaluation through comparison of TTO with sonde data outside the 1991-1992 reference period shows agreement within the uncertainty of both TOMS total ozone and the MR method (1 = 5 DU). Only data at Natal, Brazil, Brazzaville and Ascension Island are used and these are limited before 1990 and after 1996.
(2) Analysis of time-series from the 1979-92 TTO data set show seasonality generally consistent with the ozonesondes; multiple peaks in both records point to more complicated forcings than the biomass burning ozone source in determining month-to-month variations in the second half of the year (when the highest ozone occurs).
(3) The wave-one pattern in equatorial tropospheric ozone is a persistent feature, with seasonally varying amplitude and correlation with markers of dynamical variability, eg SST, prominent during ENSO events. Regional signatures of the 1982 and 1997 ENSO events appeared as localized TTO maxima in eastern South America and as minima in the eastern Pacific.
(4) From 110W to 30E, TTO shows insignificant trends from 0-12S during 1980-1990, which agrees with the findings of Kim and Newchurch [1996] for the eastern Pacific but not for eastern South America. Natal ozonesondes, the only operational site in the tropics throughout the 1980's, show no significant trend, in agreement with the TTO result.

     Comparisons between the MR method and the CCD method [Ziemke et al., 1998], which is another TOMS-only approach to retrieval of tropospheric column ozone, show very good agreement between derived stratospheric ozone and tropospheric column ozone, with the latter parameter referenced to the only independent data - Natal sondes in the 1980's. The assumptions of the two methods, as well as the selection of TOMS high-density (Level 2) data, allow extraction of stratospheric ozone from total ozone by very different approaches.

     Evaluation of the MR TTO data is an ongoing process, with EP/TOMS evaluation expanding to regions beyond the Atlantic as more tropical ozone data become available. Some of this is anticipated in the SHADOZ augmented ozonesonde network: http://code916.gsfc.nasa.gov/Data_services/shadoz. An appealing feature of the MR method is the time- averaging (2-3 maps/month) and spatial resolution, which should render them useful for process studies and field campaigns, as well as for climatological investigations of the type presented here. Readers are encouraged to use the maps on the homepage and to correspond with us about correlative ozone data.

Acknowledgments.

Thanks to M. G. Seybold and R. M. Todaro for producing the Nimbus 7 1979-1990 TTO maps and to H. Guo for the EP- and ADEOS TTO maps. A. D. Frolov helped with sonde and SAGE II analysis. Ascension Island sondes since 1997 are courtesy of F. J. Schmidlin at the GSFC/Wallops Flight Facility. Real-time TOMS processing for EP- and ADEOS-TOMS is possible thanks to the Goddard Ozone Processing Team. We appreciate discussions with J. R. Herman, P. K. Bhartia and O. Torres on TOMS absorbing aerosol data and comments on the manuscript from S. Chandra, J. R. Ziemke, R. D. McPeters and J. F. Gleason. Many thanks to S. M. Hollandsworth, T. L. Kucsera, M. G. Seybold (SM&A at NASA/Goddard) and A. V. Cresce (Univ. Maryland) for assistance with the trends program. Support for this analysis, and for some of the ozonesondes, is provided by the NASA ACMAP (Atmospheric Chemistry Modeling and Analysis Program). Additional support comes from the NASA Tropospheric Chemistry Program and an EOS Interdisciplinary Science study.

A. Application to TOMS Data Before and After 1991-1992

Figure A-1 illustrates the major features of the modified- residual method. Figure 2 in Hudson and Thompson [1998] describes the steps in deriving stratospheric column ozone, excess ozone, background tropospheric ozone and tropospheric column ozone (TTO). The fundamentals of the method are given here, for easy reference, and to clarify where the most critical assumptions are made:


(1) Although it is generally accepted that the distribution of stratospheric ozone in the tropics is not a function of longitude, there could be a latitude dependence. To remove any bias, we examined a region near 180 degrees longitude, where we assume the influence of pollution on ozone is small and that tropospheric column ozone is not a function of latitude. This means that the dependence of total ozone with latitude is assumed to be the same as the dependence of stratospheric column ozone. We used this derived dependence to normalize the total ozone at a given longitude and latitude to the value it would have at the equator for the same longitude. This permits the assumption of a single stratospheric ozone value, O₃)_str, over the latitude range in which the wave-one pattern is observed, leading to a more accurate derivation of the wave-one parameters. The data used are TOMS version 7, Level 2 data, cloud-free conditions as defined by reflectivity < 0.15, at or near 180° longitude, and where ozone is a minimum.
(2) An underlying total ozone amount (the smooth curve in Figure A-1) follows a wave-one pattern; with stratospheric and background tropospheric ozone included, this is referred to as "background total ozone," O₃)_total^back in Hudson and Thompson [1998] and Kim et al. [1996].
(3) The wave-one feature also delineates a background tropospheric ozone column amount, O₃)_tr^back, in Figure A-1. "Excess ozone," O₃)^excess, which is the amount of ozone normally ascribed to pollution [Fishman et al., 1991; Thompson et al., 1996a] is defined as follows:
O₃)^excess = O₃)_total - O₃)_total^back             
             = O₃)_total - O₃)_str - O₃)_tr^back,
where O₃)_total is total ozone from TOMS Level 2 data.
(4) Total ozone at the maximum of the background ozone (at or near 0 degrees longitude) is used to derive stratospheric ozone from:
O₃)_str = O₃)_total(0) - O₃)_tr^back(0) - O₃)^excess(0),
where O₃)_tr^back(0) is designated "A" in Figure A-1.
(5) The magnitude of the wave-one is obtained by Fourier analysis, with the wave optimized to fit along total ozone minima, as shown in Figure A-1. The assumption that the wave lies in the troposphere is signified by flat stratospheric ozone in Figure A-1.


Features (4) and (5) represent a key assumption in the modified- residual method, namely that the wave-one pattern is in the troposphere. This appears to be supported by analysis of sondes and stratospheric ozone derived from SAGE [Fishman et al., 1990; Shiotani and Hasebe, 1994], MLS and HALOE [Ziemke et al., 1996], although neither the sampling frequency of the sondes nor the accuracy of lower tropospheric profiling is definitive. Ozonesonde data have been used to show that a wave-one pattern is in the troposphere [Ziemke et al., 1996] but the sonde data used were not coincident in time.

In Hudson and Thompson [1998], O₃)_tr^back(0), which is used to obtain stratospheric ozone in (4), was derived from the wave pattern and tropospheric ozone data from sounding stations near the ozone maximum. The reason is, that for the period used in developing the MR method, just prior to and during the SAFARI/TRACE-A campaigns of 1992, no ozonesondes were launched near the tropical Pacific ozone minimum. However, at Natal, Ascension, and Brazzaville, there were more than 130 soundings during 1991-1992 [Diab et al., 1996; Nganga et al., 1996; Olson et al., 1996; Thompson et al., 1996a]. Values for each two-week period for O₃)_tr^back(N, A, B), where N, A, B represent the locations at Natal (6S, 35W), Ascension (8S, 15W) and Brazzaville (4S, 15E), respectively, were obtained through the constraint that excess ozone plus the background tropospheric ozone at each location must equal integrated tropospheric ozone from the ozone sondes:



O₃)^excess(N, A, B) + O₃)_tr^back(N, A, B) = O₃)^sonde(N, A, B)
Averaging of O₃)_tr^back(N, A, B) over the three sites was used to obtain O₃)_tr^back(0) from the longitudinal dependence of the wave. The relationship in (4) gave stratospheric ozone, O₃)_str, by subtraction from total ozone. When the results for O₃)_tr^back obtained from the three sounding sites were averaged, a seasonally varying signal for O₃)_tr^back(0) could be fit with a sinusoidal function: O₃)_tr^back(0, t), where "t" refers to time (Figure 5 in Hudson and Thompson, 1998). Our best estimate of O₃)_tr at each latitude/longitude point in the 1991-1992 period, referred to as "TTO", was obtained using this function and the above procedure.

Preliminary extension of the 1991-1992 background tropospheric ozone signal at the wave maximum, O₃)_tr^back(0, t), to TOMS total O₃ data from 1990 gave TTO in good agreement with tropospheric ozone measured by ballooon-borne sondes at Ascension, Natal, and Brazzaville Figure A-2(A),(B),(C). Hence, it was decided to use the MR method with the 1991-1992 O₃)_tr^back(0, t) function during the EP and ADEOS periods (1996-present) and backward through the Nimbus 7 era to give two time-series of TTO.

For each 13-15-day period over which version 7 total ozone from Nimbus 7 is averaged, the procedure used is identical to that described in Hudson and Thompson [1998]. Normalization for total ozone prescribes the location of the ozone minimum and permits derivation of a single O₃)_str, subject to the magnitude and geographical boundaries of the wave-one pattern. The first set of N7/TTO maps was posted as gif images on this homepage in April 1998; reprocessing took place in January 1999.

Hudson and Thompson [1998] and Kim et al. [1996] presented a thorough error analysis of TTO. Some of the imprecisions are due to assumptions made in adopting a seasonally varying background tropospheric ozone column near the ozone maximum: 1 = 4.8 DU (Table 1 in Hudson and Thompson, 1998). Since then, we have made tests using radiances calculated from actual ozone profiles. These indicate that the standard deviation of "atmospheric noise" due to profile errors is about 5 DU in the tropics. This is similar to the 1- deviation (5.3 DU) between individual ozonesondes and the 2-week averages used in referencing O₃)_tr^back(0) to the sondes (Table 4 in Hudson and Thompson, 1998). The latter value is well within experimental error of both TOMS and the sondes, especially when one considers the single-point sonde measurement and the 50x50-km averaging in TOMS. This number is not to be equated with the error in the TTO method (for example, that which could come from derivation of the wave-one constants), which Hudson and Thompson [1998] estimate to be ± 5 DU. Note that a 5-DU error may be less than 10% of tropospheric column during the seasonal Atlantic tropospheric O₃ maximum or localized pollution events in other regions. It is 25% near the Pacific tropospheric ozone minimum, which averages 20 DU over the course of a year.

Chandra, S., J. R. Ziemke, W. Min and W. G. Read, Effects of 1997-1998 El Nino
     on tropospheric ozone and water vapor, Geophys. Res., Lett., 25, 3867-
     3870, 1998.
Chandra, S., J. R. Ziemke, and R. W. Stewart, An 11-year solar cycle in
     tropospheric ozone from TOMS measurements, Geophys. Res. Lett., 26, 185-
     188, 1999.
Chatfield, R. B., J. A. Vastano, H. B. Singh, and G. Sachse, A general model
     of how fire emissions and chemistry produce African/oceanic plumes (O3,
     CO, PAN, smoke) in TRACE A, J. Geophys. Res., 101, 24279-24306, 1996.
Diab, R. D., et al., Vertical ozone distribution over southern Africa and
     adjacent oceans during SAFARI-92, J. Geophys. Res., 101, 23809-23821,
     1996.
Fishman, J. and V. G. Brackett, The climatological distribution of
     tropospheric ozone derived from satellite measurements using version 7
     Total Ozone Mapping Spectrometer and Stratospheric Aerosol and Gas
     Experiment data set, J. Geophys. Res., 102, 19275-19278, 1997.
Fishman, J., C. E. Watson, J. C. Larsen, and J. A. Logan, The distribution of
     tropospheric ozone determined from satellite data, 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, E. V. Browell, and W. B. Grant, Tropospheric
     ozone derived from TOMS/SBUV measurements during TRACE-A, J. Geophys.
     Res., 101, 24069-24082, 1996.
Fujiwara, M., K. Kita, S. Kawakami, T. Ogawa, N. Komala, S. Saraspriya and A.
     Suripto, Tropospheric ozone enhancements during the Indonesian forest
     fire event in 1994 and 1997 as revealed by ground-based observations,
     Geophys. Res Lett., In Press, 1999.
Hollandsworth, S. M., R. D. McPeters, L. E. Flynn, W. Planet, A. J. Miller,
     and S. Chandra, Ozone trends deduced from combined Nimbus 7 SBUV and
     NOAA 11 BUV/2 data, Geophys. Res. Lett., 22, 905-908, 1995.  
Hsu, C. N., J. R. Herman, O. Torres, B. N. Holben, D. Tanre, T. F. Eck, A.
     Smirnov, B. Chatenet, and F. Lavenu, Comparisons of the TOMS aerosol
     index with Sun-photometer aerosol optical thickness: Results and
     applications, J. Geophys. Res., 104, 6269-6279, 1999.
Hudson, R. D., J. Kim, and A. M. Thompson, On the derivation of tropospheric
     column ozone from radiances measured by the Total Ozone Mapping
     Spectrometer, J. Geophys. Res., 100, 11138-11145, 1995.
Hudson, R. D., M-H. Ahn, and A. D. Frolov, An evaluation of TOMS-based methods
     for retrieving tropospheric ozone, CACGP presentation, 1998.
Jiang, Y. and Y L. Yung, Concentration of tropospheric ozone from 1979 to 1992
     over tropical Pacific South American from TOMS Data, Science, 272, 714-
     716, 1996.
Kim, J-H., and M. J. Newchurch, Climatology and trends of tropospheric ozone
     over the eastern Pacific Ocean: The influences of biomass burning and
     tropospheric dynamics, Geophys. Res. Lett., 23, 3723-3726, 1996.
Kim, J. H., and M. J. Newchurch, Biomass-burning influence on tropospheric
     ozone over New Guinea and South America, J. Geophys. Res., 103, 1455-
     1461, 1998.
Kim, J-H., R. D. Hudson and A. M. Thompson, A new method of deriving time-
     averaged tropospheric column ozone over the tropics using total ozone
     mapping spectrometer (TOMS) radiances: Intercomparison and analysis
     using TRACE-A data, J. Geophys. Res., 101, 24317-24330, 1996.
Kirchhoff, V. W. J. H., R. A. Barnes, and A. L. Torres, Ozone climatology at
     Natal, Brazil, from in-situ ozonesonde data, J. Geophys. Res., 96,
     10899-10909, 1991.
Kirchhoff, V. W. J. H., J. R. Alves, F. R. da Silva and J. Fishman,
     Observations of ozone concentrations in the Brazilian cerrado during the
     TRACE-A field expedition, J. Geophys. Res., 101, 24,029-24,042, 1996.
Kley, D., P. J. Crutzen, H. G. J. Smit, H. Vömel, S. J. Oltmans, H. Grassl,
     and V. Ramanathan, Observations of near-zero ozone concentrations over
     the convective Pacific: Effects on air chemistry, Science, 274, 230-233,
     1996.
Komhyr, W. D., P. R. Oltmans, P. R. Franchois, W.F.J. Evans, and W. A.
     Matthews, The latitudinal distribution of ozone to 35 Km altitude from
     ECC ozonesonde observations, 1985-1987, Ozone in the Atmosphere, ed.
     R.D. Bojkov and P. Fabian, Proceedings of Quadrennial Ozone Symposium
     1988 and Tropospheric Ozone Workshop, pp. 147-150, 1989.    
Krishnamurti, T. N., M. C. Sinha, M. Kanamitsu, D. Oosterhof, H. Fuelberg, R.
     Chatfield, D. J. Jacob, and J. Logan, Passive tracer transports relevant
     to the TRACE-A experiment, J. Geophys. Res., 101, 23889-23908, 1996.
Logan, J.A., Trends in the vertical distribution of ozone: An analysis of
     ozone sonde data, J. Geophys. Res., 99, 25553 25585, 1994.
Nganga, D., A. Minga, B. Cros, C. Bouka Biona, J. Fishman and W. B. Grant, The
     vertical distribution of ozone measured at Brazzaville, Congo, during
     TRACE-A, J. Geophys. Res., 101, 24095-24103, 1996.          
Novelli, P. C., K. A. Masarie, and P. M. Lang, Distribution and trends of
     carbon monoxide in the lower troposphere, J. Geophys. Res., 103, 19015-
     19033, 1998.
Olson, J. R., J. Fishman, V. W. J. H. Kirchhoff, D. Nganga, and B. Cros, An
     analysis of the distribution of O3 over the southern Atlantic region, J.
     Geophys. Res., 101, 24083-24094, 1996.
Piotrowicz, S. R., H. F. Bezdek, G. R. Harvey, M. Springer-Young, and K. J.
     Hanson, On the ozone minimum over the equatorial Pacific Ocean, J.
     Geophys. Res., 96, 18679-18687, 1991.
Shiotani, M., Annual, quasi-biennial and El Nino-Southern Oscillation (ENSO)
     time-scale variations in Equatorial total ozone, J. Geophys. Res., 97,
     7625-7634, 1992.
Shiotani, M and F. Hasebe, Stratospheric ozone variations in the equatorial
     region as seen in Stratospheric Aerosol and Gas Experiment data, J.
     Geophys. Res., 99, 14,575-14,584, 1994.
Stolarski, R. S., P. Bloomfield, R. McPeters, and J. Herman, Total ozone
     trends deduced from Nimbus 7 TOMS data, Geophys. Res. Lett., 18,
     1015-1018, 1991.
Thompson, A. M., et al., Ozone observations and a model of marine boundary
     layer photochemistry during SAGA-3, J. Geophys. Res., 98, 16955-16968,
     1993.
Thompson, A. M., K. E. Pickering, D. P. McNamara, M. R. Schoeberl, R. D.
     Hudson, J. H. Kim, E. V. Browell, V. W. J. H. Kirchhoff, and D. Nganga,
     Where did tropospheric ozone over southern Africa and the tropical
     Atlantic come from in October 1992? Insights from TOMS, GTE/TRACE-A and
     SAFARI-92, J. Geophys. Res., 101, 24251-24278, 1996a.
Thompson, A. M., et al., Ozone over southern Africa during SAFARI-92/ TRACE-A,
     J. Geophys. Res., 101, 23793-23807, 1996b.
Thompson, A. M., R. D. Hudson and H. Guo, A TOMS-bsed view of tropical
     tropospheric ozone (TTO) from large fires associated with the 1997-1998
     El Nino, Eos, Trans. AGU, Fall Meeting, 1998.
UNEP, WMO Global Ozone Research and Monitoring Project, Report No. 44,
     Scientific Assessment of Ozone Depletion: 1998, Chapter 8, Tropospheric
     ozone and related processes, by J. Lelieveld and A. M. Thompson, Geneva,
     1998.
WMO (World Meteorological Organization), SPARC/IOC/GAW Assessment of Trends in
     the Vertical Distribution of Ozone, ed. By N. Harris, R. Hudson and C.
     Phillips, SPARC Report No. 1, WMO Global Ozone Research and Monitoring
     Project, Report No. 43, Geneva, 1998.
Ziemke, J. R. and S. Chandra, Comment on "Tropospheric ozone derived from
     TOMS/SBUV measurements during TRACE A," by J. Fishman et al., J.
     Geophys. Res., 103, 13903-13906, 1998.
Ziemke, J. R., S. Chandra, A. M. Thompson and D. P. McNamara, Zonal
     asymmetries in southern hemisphere column ozone: Implications of biomass
     burning, J. Geophys. Res., 101, 14421-14427, 1996.
Ziemke, J. R., S. Chandra, and P. K. Bhartia, Two new methods for deriving
     tropospheric column ozone from TOMS measurements: The assimilated UARS
     MLS/HALOE and convective-cloud differential techniques, J. Geophys.
     Res., 103, 22115-22128, 1998.

Plate 1. (A) Typical Nimbus 7 period map from TTO website. URL = (http://metosrv2.umd.edu/~tropo). Each month's record consists of two maps. The first is based on averaging Days 1-15 Level 2, low- reflectivity TOMS total ozone; second image is based on averaging Day 16 to the end of the month. Processing with the MR method extends from 20°N to 30°S but maps are restricted to the range of the wave-one pattern. The TTO for October 1992 agrees with tropospheric ozone from sondes during the SAFARI-92 and TRACE-A campaigns [Hudson and Thompson, 1998]; (B) same as (A) except for an ADEOS map during the southern hemisphere 1996 burning season. Real-time maps show 10°S-10°N. The only ozonesonde station for which data are available in September 1996 are just south of this map. At Samoa, 14°S, 171°W, integrated tropospheric ozone from soundings was 30-44 DU; TTO at 10°S, 171°W is 35 DU. Fig. 1 Comparison of derived stratospheric ozone with SAGE II ozone, averaged over 10°S-10°N, and recorded from 1985-1991. Derived stratospheric ozone from MR method is based on averaging over O₃)_str determined for latitudes within the 10°N-10°S band, using latitudinal adjustment shown in Figure 3 in Hudson and Thompson [1998]. From June 1991-1993, stratospheric aerosols from the Mt Pinatubo volcanic eruption made reliable SAGE ozone retrievals impractical. Nimbus 7/TOMS derived stratospheric ozone record compared with the UARS/MLS from September 1991-December 1992 appears in Figure 6 of [Hudson and Thompson, 1998]. Fig. 2 Comparison of derived stratospheric ozone with UARS MLS stratospheric column ozone, averaged over 10°S-10°N, 1996-1998. Same method for deriving stratospheric ozone as in Figure 3. For MLS, integration of ozone is from 1-100 hPA; below 46 hPa, MLS ozone precision is 50%. Fig. 3 Comparison of derived stratospheric ozone, O₃)_str appropriate for 8S (latitudinal adjustment as in Fig. 3, Hudson and Thompson, 1998), solid line, with stratospheric ozone determined by subtracting integrated tropospheric ozone at Ascension (8°S, 15°W) from TOMS total ozone. Fig. 4 Derived stratospheric ozone, O₃)_str, appropriate for 6°S (latitudinal adjustment as in Fig. 3, Hudson and Thompson, 1998), solid line, 1979- 1992, at Natal, Brazil (6°S, 35°W). Natal is the only tropical sounding station with regular sondes since 1978 [Kirchhoff et al., 1991; 1996]. * denotes stratospheric ozone computed by subtracting twice-per-month averaged tropospheric ozone from Natal sondes from TOMS total ozone. Stratospheric ozone from the sondes is not an exact value because some extrapolation above balloon burst must be assumed. Fig. 5 Comparison of integrated tropospheric ozone from ozonesondes with TTO from the modified-residual method for 1979-1992 at Natal, Brazil (6°S, 35°W). Natal is the only tropical sounding station with sondes since 1978 [Kirchhoff et al., 1991; 1996]. Line denotes twice-per-month averaged TTO from 5-7°S and 32-38°W; * symbol is integrated ozone from the sounding, surface to 100 hPa. Deviation of TTO from sonde value appears at bottom of figure (). Fig. 6 Comparison of TTO from Earth-Probe/TOMS for late July 1997-May 1998 at Ascension (8°S, 15°W). Ascension soundings were re-activated in 1997 after a 5-year hiatus; launch frequency in 1997-1998 is twice per week. Line denotes twice-weekly averaged TTO and * symbol is integrated ozone from the sounding, surface to 100 hPa. Deviation of TTO from sonde is . Fig. 7 Same as Figs. 5 and 6 except that comparison is between tropospheric ozone from Samoan sondes (14°S, 171°W) and N7/TTO obtained by extending processing to 14°S, which may be south of the wave-one pattern. Fig. 8 TTO derived from the MR method over the Nimbus 7 period at three ozone sounding sites, for which latitude-longitude given in caption for Figure A-2 A = Natal; B = Ascension; C = Brazzaville). Latitude is 0-12°S for the other regions, with longitudes as follows: D = Eastern Pacific, 80- 110°W; E = eastern So. America, 40-70°W; F = south Atlantic, 0-40°W; G = southern Africa, 0-30°E. Deseasonalized mean, determined by linear- regression model, is the value at the beginning of each series. Fig. 9 Model determined seasonality for Natal (A), Ascension (B) and Brazzaville (C) where the solid line signifies analysis of 14-year TTO corresponding to each site and dashed line refers to seasonality based on integrated tropospheric ozone from the sonde record: 1979-1992 at Natal, 1990-1992 for Brazzaville. Model was not applied to Ascension because of large data gaps in first half of the year. Fig. 10 Amplitude of tropospheric wave, in DU (normalized to mean, outer scale, open circles), over 14-year N7/TOMS record [cf Ziemke et al., 1998]. Pattern of minima twice per year (in May-June and December- January) represents N-S transition in Intertropical Convergence Zone (ITCZ). Stable ITCZ is associated with convective transport causing dilution of the tropospheric ozone column over the Pacific relative to the Atlantic [Piotrowicz et al., 1991]. Effect of 1982-83 and 1987 ENSO events on wave amplitude is illustrated by correlation of wave amplitude with sea-surface temperature anomaly (filled circles, SST, deg K). Positive anomaly signifies more convection over the Pacific, a greater Atlantic-Pacific ozone contrast, ie larger wave amplitude. Fig. 11 (A) Time-series of TTO (solid line) at Natal (6°S, 35°W), with best-fit (----) linear trend assuming seasonal coefficients, assumed over 1980-1990. (B) Same as (A) except that the Natal ozonesonde record, with sondes averaged to twice-per-month frequency, is basis for model analysis. Fig. 12 (A) Ascension Island TTO for the whole calendar years of the Nimbus 7 period, 1980-1990. As in Fig. 11, model best-fit (dashed line) and deseasonalized trend (straight line) are given. (B), (C), (D). Same as A for eastern South America (0-12°S, 40-70°W); southern Atlantic (0-12°S, 0- 30°E); east southern Pacific (0-12°S, 80-110°W). Southern African TTO (0- 12°S, 0-30°E), not shown, is similar to the south Atlantic. Fig. A-1 Schematic of modified-residual (MR) method for deriving tropical tropospheric ozone and stratospheric ozone, given total ozone from TOMS, O₃)_total, a wave-one pattern (with amplitude ) and excess ozone, O₃)^excess. Basis of distinguishing stratospheric and tropospheric ozone is a 2-year climatology of ozonesondes at three sites near the wave maximum (0° longitude). A fixed, seasonally varying O₃)_tr^back(0) = A, is assumed to apply over the duration of Nimbus 7 (1979-1992), Earth-Probe (Aug. 1996- present) and ADEOS (Sept. 1996 - April 1997) TOMS. Fig. A-2 Comparison of tropical tropospheric ozone (TTO) derived from the modified-residual method for 1990 and 1991 at (A) Ascension Island (8°S, 15°W); (B) Brazzaville, Congo (4°S, 15°E) (C) Natal, Brazil (6°S, 35°W). Line denotes twice-per-month averaged TTO from Nimbus 7/TOMS and * symbol is integrated ozone from the sounding, surface to 100 hPa. The precision of the MR method is ± 5 DU (Dobson Units, 1 DU = 2.69 x 10(16) cm-2). This is derived from averaging tropospheric ozone over 1-2-week periods and from inherent imprecision in the TOMS total ozone retrieval.