Lower Tropospheric Ozone 

Using TOMS Version-7 Level-2 clear-sky (reflectivity ?20%) ozone measurements corrected for aerosol effects and sea-glint errors, we derived Lower Tropospheric Ozone (LTO) west and east of the Andes, the Mexican and Rocky Mountains, the mountains in Africa and the Arabian Peninsula, New Guinea, and the Himalayan Mountains. Figure 1 shows a map of locations where we derive LTO.

1. Methodology

We derive the lower tropospheric ozone using the topographic contrast method, i.e. subtracting total column ozone over the high elevation mountains from the ozone over adjacent sea level surface to obtain the ozone from sea level to mountain top. Different from previous studies by Jiang and Yung [1996], Kim and Newchurch [1996, 1998], we use TOMS high-resolution Level-2 clear sky measurements to avoid ozone retrieval errors associated with clouds and correct sun glint and tropospheric aerosol effects as well.

We define mountain regions with terrain pressure less than 800 mb between 40?S-40?N latitude. The daily average TCO at mountain levels results from using all available measurements with terrain pressure less than 800 mb within ??longitude and 1?latitude of the mountain peak. On both east and west sides of mountains, we obtained the corresponding TCO using measurements with terrain pressure greater than 950 mb within ?7.5?longitude. To compare the magnitude of derived LTO across all locations, it is convenient to express the LTO in mixing ratio rather than ozone column. Assuming a well-mixed LTO and hydrostatic atmosphere, we could calculate the LTO mixing ratio using the difference between TCO at the mountain level (ML) and at the sea level (SL):

TCO(s) and TCO(m) are the daily average TCO (DU) at SL and ML, respectively; P(s) and P(m) are the average terrain pressure (atm) at SL and ML, respectively. The mountain terrain pressures at these locations range from 800 mb to 530 mb, with an average of 720 mb (about 3 km in height). Due to requiring only clear-sky measurements, we might not obtain the daily ozone mixing ratio every day. In that case, the average of TCO at ML or SL within ??latitude and ? days provides the ozone mixing ratio.

An important assumption in the TCM is that both stratospheric ozone and upper-tropospheric ozone (defined here as the ozone amount between the ML and the tropopause) are the same over both the mountain and the adjacent land (or ocean). The derived daily LTO shows large variability from day to day, and some ozone mixing ratios are negative or extremely large, suggesting the assumption might not be correct on a daily bias. To reduce this variability on the derived LTO, we formed the monthly mean ozone mixing ratio by requiring at least 10 daily values from 15?S-15?N and at least 15 values at the other latitudes. Because the natural variability of ozone is about 30% of the monthly mean in middle latitudes, and smaller in tropical regions, we omit those monthly mean values with relative standard error greater than 30% or with standard deviations greater than 10 ppbv.

We compare the derived LTO with the ozonesonde observations at Boulder (from CMDL), Cristobal, Fiji, and Samoa (from SHADOZ) in Figure 2. The derived ozone mixing ratios east of Rocky are limited to the summer season because of the variability constraint. We can see that the derived LTO agrees very well with the annual variability of the Boulder ozonesonde observations. The derived LTO west of Andes (0-2S), Andes (18S-20S), East of New Guinea (5S-7s) agree very well with the seasonality of SHADOZ-measured LTO at Cristobal, Samoa, and Fiji, respectively. The strong similarity in seasonal behavior strengthens our confidence in using the TCM method to analyze the seasonality and trends of LTO at the selected mountainous regions.

3. Seasonal variation of lower tropospheric ozone

Figure 3 shows the 18-year (1979-1992 N7 TOMS, 1996-1999 EP TOMS) average monthly mean lower tropospheric ozone at the above mentioned regions. To investigate the influence of biomass burning on enhanced tropospheric ozone, we present the monthly mean fire count map (2 longitude by 2 latitude) derived from ATSR-2 night-images in four seasons in 1998 in Figure 4 (from http://shark1.esrin.esa.it/ionia/FIRE/AF/ATSR/) with superimposed LTO locations from Figure 1.

The influence of biomass burning is evident in the seasonal variation of LTO on both sides of the Andes Mountains between 23S-2 N, in southern Africa, in southern Sudan, and, especially in relation to the El Nino modulation, west of New Guinea. Also, in northern equatorial Africa consistent with the seasonality of the measured fires, we see a seasonal cycle in LTO also observed by one other satellite technique.

Similar to the ozone in the northern tropics of South America, little seasonal variation is found in Kenya (1 S -2 N), consistent with the Nairobi ozonesonde record . In North Sudan (15-20N), where it is dry throughout the year and there is almost no biomass burning, we see a summer maximum and less annual variation than in the biomass-burning regions. High ozone amounts with a summer maximum are also seen in the Iraq region (30 -35 N), and we speculate that these high amounts are due to precursors from the oil fields. The importance of stratospheric intrusion is evident in the seasonality observed in the eastern Pacific Ocean at 15 -30 N and 30-20S, and east of the Himalayas in western China.

In the continental regions such as west and east of the Rocky Mountains (35-40N), east of the Mexican Mountains (23-30N), Iraq, and western China, which are subject to the anthropogenic emissions of ozone precursors, high concentrations of ozone are usually found in a summer maximum. In coastal regions such as in the eastern Pacific Ocean west of the Andes and Mexican mountains, an ozone minimum usually occurs in the summer, suggesting photochemical ozone destruction in low NOx and high H₂O environment. Precipitation also plays an important role in ozone loss due to stronger wet deposition, convective activity, and a high H₂O environment such as the case in South Sudan.

4.  Trends of Lower Tropospheric Ozone

Because of an unresolved bias between the LTO derived from N7 and EP TOMS data due to an N7/EP offset, we use LTO derived from only N7 TOMS data to examine ozone trends. After removing the seasonal cycle in LTO we express the deseasonalized ozone linear-regression trends (Figure 5) in percent deviations from the monthly means with ?5% confidence intervals for months with at least 7 of the possible 14 values. We also calculate seasonal trends for each season using the series of deseasonalized ozone for the appropriate months.

We find positive trends significant at the 95% confidence level between 23-12S west of the Andes (0.7?.4 %/yr), and 1S-8N in Kenya and Somalia coastal regions (1.9?.4 %/yr). We find significant negative trends between 13-6S east of the Andes (–0.8?.7 %/yr). Our calculated trends from the Level-2 NIMBUS-7 TOMS data are consistent with the eastern Pacific results of and , studies which used Level-3 NIMBUS-7 TOMS data. Although the latitudinal structure of the trends on the east side of the Andes is similar to the structure on the west side, the trends are more negative on the east side. We find zero trends between 26-13S east of the Andes, and significant negative trends at 13-6S. The corresponding trends on the west side are positive between 26-13S and zero between 13-6S. The large positive trends in Kenya and Somalia are mainly in the fall and summer (i.e., the minimum ozone season). LTO west of New Guinea 5.5S also shows positive trends but with a larger standard deviation (1.4?.5 %/year), and this positive trend also occurs mainly in the minimum ozone season. These positive trends in the minimum ozone season indicate an increase in background ozone levels.