Abstract. Data from a 63-station radiosonde network are used to estimate the variation in global tropospheric (850-300 mb) temperature between 1958 and 1989. The annual temperature was a maximum in 1988, 0.42°C above the 1958-88 mean. However, the 1989 value is indicated to be only 0.12°C above this mean. During 1958-88, there has been a correlation of 0.76 (significant at the 0.1 level) between annual values of global tropospheric-temperature deviation and annual values of sea-surface temperature (SST) deviation in eastern equatorial Pacific two seasons earlier. The associated linearregression line indicates that an annual SST deviation of 1°C in the region 12°S-2°N, 180°90°W has been related on average to an annual global tropospheric-temperature deviation of 0.36°C. The annual values of global tropospheric temperature have been adjusted based on this regression. With this adjustment, the year-to-year variability in global tropospheric temperature is halved, the increase in decadalmean temperature between the 1960's and 1980's is reduced from 0.33°C to 0.24°C, the annual temperature is a maximum in 1989 (0.39°C above the 31-year mean) rather than in 1988, and there is much more convincing evidence that the eruptions of Agung in 1963 and El Chichon in 1982 decreased global tropospheric temperatures by 0.2-0.3°C for about 3 years.

Introduction

With the acquisition of temperature data for 1989, 32 years of data have become available for the 63 radiosonde stations used by Angell and Korshover (1983) to estimate global temperature variations in troposphere and low stratosphere. This paper considers only annual temperature variations in the tropospheric 850-300 mb layer for the interval 1958-89. In order to reduce the interannual variability in the record, adjustments are made for the influence of El Nino and anti El Nino on global tropospheric temperature. The procedure is similar to that of Jones [1989], who regressed global and hemispheric surface temperatures onto the Tahiti-Darwin Southern Oscillation index. is shown that through the use of this procedure temperature trends become more apparent and there is a better resolution of volcanic effects. Discussed briefly is whether the removal of El Nino and anti El Nino influences on global tropospheric temperature results in a more or less representative estimate of longterm temperature changes associated with the greenhouse effect.

It

Procedures

At each of the 63 radiosonde sites [see Figure 1 of Angell and Korshover, 1983, for site locations] mean-monthly geopotential heights have been obtained for 850 and 300 mb mandatory pressure surfaces using both daily teletype data collected at the National Meteorological Center of the National Weather Service, NOAA, and data published in "Monthly Climatic Data for the World" by the National Environmental Satellite, AccordData, and Information Service of NOAA. ing to the hydrostatic equation, the difference in height of two pressure surfaces (thickness) is proportional to the mean virtual temperature of the layer between these pressure surfaces. At each station the mean-monthly heights have been averaged by season (December-JanuaryFebruary, etc.) and a mean seasonal temperature determined for the 850-300 mb (tropospheric) layer. Seasonal temperature deviations from the long-term means were then evaluated at each station, the average of the four seasonal deviations yielding the annual deviation used herein.

The annual temperature deviations at the individual stations have been averaged (with equal weighting) to obtain annual temperature deviations for north and south polar (60°-90°), north and south temperate (30°-60°), north and south subtropic (10°-30°) and equatorial (10°S10°N) climatic zones. A 1,2,2,1 weighting of the temperature deviations for polar, temperate, subtropical and equatorial zones, respectively (roughly proportional to their areal extent), defines the annual temperature deviation for the hemisphere, and the average of the deviations for the two hemispheres defines the annual temperature deviation for the globe.

Temperature Record

Based on the 63-station radiosonde network, Figure 1 shows the variation in global annual temperature in the tropospheric 850-300 mb layer between 1958 and 1989, expressed as deviations from the 1958-88 mean. Since the annual deviation is determined from an average of the 4 seasonal deviations, December of the previous year contributes to the value for the year in question. Also indicated in Figure 1 are the powerful volcanic eruptions of Agung (8°S) in 1963 and El Chichon (17°N) in 1982.

Following a fairly uniform decrease in global tropospheric temperature between 1959 and 1965 [see also Starr and Oort, 1973], there has been an irregular warming, culminating in a record high temperature in 1988 (0.42°C above the 195888 average). The five warmest years of this 32year record are in the 1980's, namely, 1980, 1981, 1983, 1987 and 1988 [see also Hansen and Lebedeff, 1988; Jones et al., 1988), although in

Variation of global temperature in the tropospheric 850-300 mb layer, 1958-89. The annual values are expressed as deviations (°C) from the 1958-88 mean.

this analysis the temperatures in 1980 and 1981 barely exceed the temperatures in 1959 and 1973. There is little evidence from Figure 1 of a unique cooling of the global troposphere after the eruptions of Agung and El Chichon. Note the indicated decrease in global tropospheric temperature in 1989 to a value only 0.12°C above the 1958-88 mean.

In Figure 1 the average year-to-year change (without regard to sign) in global tropospheric temperature is 0.20°C. This is an appreciable value in the context of both the 0.5°C warming of the global surface observed during the last century [e.g., Folland et al., 1984; Hansen and Lebedeff, 1988; Jones et al., 1988] and the temperature changes to be expected during the early stages of a greenhouse effect. It is desirable to reduce this interannual variability, thereby enabling the signal of any longterm temperature change to stand out more clearly from the noise of the interannual variability. A reduction in this interannual variability is the main purpose of this paper.

SST Influence

There has been a close relation between seasurface temperature (SST) in eastern equatorial Pacific (the El Nino region) and tropospheric temperature in the tropics [Newell and Weare, 1976; Angell, 1981; Navato et al., 1981; Angell and Korshover, 1983; Pan and Oort, 1983; Angell, 1988]. It is widely believed that most of the effect of increased SST on atmospheric temperature is via the altered pattern of convection (diabatic heating) and does not result from sensible heating. That is, an above-average SST in equatorial Pacific leads to increased moist convection and the release of above-average amounts of latent heat of condensation. Organized convection appears to play a very important role in heating the tropical troposphere.

Because the tropics comprise half the globe, there is also a good relation between this equatorial SST and global tropospheric temperature (Angell, 1988, Figure 2]. In both cases the tropospheric temperature is indicated to lag the equatorial SST by about two seasons. The

reason for the 2-season lag is still uncertain, the atmosphere not being expected to hold latent heat for that long. It has been suggested by a reviewer that because organized convection appears to require a threshold SST, it is a combination of the annual cycle of tropical Pacific SST and anomaly that is important, giving rise to the largest total SST at a time in the annual cycle which may not be the same as the time of anomaly maximum itself, particularly in the index region used. While the 2-season lag has been incorporated in the subsequent analysis, it is admittedly a disquieting feature of the analysis because of the lack of physical understanding of the reason for such a large lag. As a recent example of the relation between global tropospheric temperature and equatorial SST, Figure 1 shows that, based on this record, there was an 0.30°C decrease in global tropospheric temperature between 1988 and 1989. Figure 2 indicates how the tropospheric temperature changes between the two years varied by climatic zone. There was a large temperature decrease in the tropics, a modest decrease in both polar zones, but a slight temperature increase in both temperate zones. The large temperature decrease in the tropics is related to the rapid transition from a very warm SST (El Nino) in eastern equatorial Pacific in 1987 and early 1988 to an unusually cold SST (anti El Nino) in this region in late 1988 and early 1989 [e.g., Trenberth et al., 1988; Angell, 1990, Figure 11]. The large decrease in tropical tropospheric temperature overwhelms the tenperature changes in the other climatic zones, resulting in the 0.30°C decrease in global temperature between 1988 and 1989 in Figure 1. If this SST influence on tropical temperature, and thereby on global temperature, could be removed from the record, the interannual temperature variability would be considerably reduced.

The dots in Figure 3 show, for years 1958-88, the relation between annual values of global tropospheric temperature deviation (see Figure 1) and annual values of SST deviations from the mean in the region 12°S-2′′N, 180°-90°W, the latter kindly made available by H. Diaz of NOAA from the COADS data set. This region is of slightly greater latitudinal extent than that

Fig. 3.

-12

1989 O

12°S-2°N, 180°-90°W SST (°C)

Comparison of global annual temperature deviations in the tropospheric 850-300 mb layer with annual sea-surface temperature (SST) deviations in eastern equatorial Pacific (12°S2°N, 180°-90°W). The annual SST is evaluated for a year beginning two seasons earlier than the tropospheric-temperature year. The dots represent the annual comparisons for the years 1958-88 on which the correlation of 0.76 and regression-line slope of 0.36 are based. The small circle indicates the comparison for the tropospheric-temperature year of 1989.

1095

2°N, 180°-90°W was associated on average with an annual global tropospheric-temperature deviation of 0.36°C two seasons later. Because of the significance of the relation, it is reasonable to attempt to remove the SST influence on global temperature by means of this regression line.

Adjusted Temperature Record

The annual temperature deviations in Figure 1 have been adjusted based on the linear regression line of Figure 3, and are shown in Figure 4. As an example of the adjustment procedure, in Figure 1 the global tropospheric temperature in 1989 is indicated to be 0.12°C above the 1958-88 mean. However, the equatorial SST for the year displaced two seasons earlier was 0.76°C below the 1958-88 mean (strong anti El Nino). Since a -1.0°C SST deviation has been associated on average with a -0.36°C global temperature deviation (see Figure 3), an SST deviation of -0.76°C would be associated with The a -0.27°C global temperature deviation. difference between -0.27°C and the observed 0.12°C in Figure 1 is the 0.39°C shown in Figure 4. The global tropospheric temperature in 1989 was not nearly as cold as would have been expected from the previous record and the strength of the anti El Nino in late 1988 and early 1989, as is apparent from the anomalous location of the small circle in Figure 3.

GLOBE, 850-300 MB

used previously [e.g., Angell, 1988), and although because of its hemispheric asymmetry may not be the optimum region for detection of the El Nino-Southern Oscillation (ENSO) signal, the SST data for this region were already available and did not have the continuity problems of the data set used previously. It is emphasized that this region was not chosen to maximize the SST correlation with global tropospheric tenperature. Because of the 2-season lag between equatorial SST and global tropospheric temperature, the SST data are for a year beginning 2 seasons earlier than the tropospherictemperature year. For example, the annual tropospheric temperature for 1989 is compared to the SST for the year extending from the northern summer of 1988 to the northern summer of 1989. The three dots furthest below the regression line in Figure 3 represent the years 1964, 1965 and 1966 following the Agung eruption, and indicate an anomalous coolness following the eruption. The year 1989 (represented by the small circle) is furthest above the regression line and is considered in the next section.

The correlation between the 31 annual values (dots) of equatorial SST and global tropospheric temperature lagged in the above manner is 0.76. Because of the negligible serial correlation in annual SST (lag 1 autocorrelation of 0.11), this correlation of 0.76 is significant at the 4-sigma (0.1) level based on Fisher's 2 test [Hoel, 1947]. Thus, 55-60% of the annual variance of global tropospheric temperature has been accounted for by ENSO-related effects. This is to be compared with a value of 30-35% found by Jones [1989] for a much longer time period, using surface temperatures. The slope of the linear-regression line in Figure 3 indicates that during this 31-year interval an annual SST deviation of 1°C in the region 12°S

1096

of ENSO activity can be considered independent of global temperature change, this emphasizes the need for extreme caution in using the temperatures of particular years as evidence either for or against a greenhouse effect.

2. There is now much more convincing evidence that the eruptions of Agung in 1963 and El Chichon in 1982 decreased the global tropospheric temperature, apparently by 0.2-0.3°C for about 3 years. As an example of the possible impact of the adjustments on estimates of volcanic effect, in Figure 1 the unadjusted 1983 temperature deviation is 0.40°C one year after El Chichon, whereas in Figure 4 the adjusted 1983 temperature deviation is -0.01°C!

3. The average year-to-year change in global tropospheric temperature has been reduced by 50% (from 0.20°C to 0.10°C), and the variance of the annual temperature has been reduced by nearly two-thirds, facilitating the detection of decadal variations. The reduction in variance is most apparent in the 1970's.

4. The increase in global tropospheric temperature between the decade of the 1960's and the decade of the 1980's is reduced from 0.33°C before adjustment to 0.24°C after adjustment, indicating in part the influence of the strong El Nino's of 1982-83 and 1987-88 on global tropospheric temperature even from a decadal point of view.

Discussion

The question arises as to how ENSO activity might fit into any greenhouse scenario, and whether the removal of the ENSO influence on global temperature, attempted here, results in a more or less representative estimate of the long-term temperature changes associated with such a scenario. In the index region used there has been an increase in SST of about 0.2°C between the 1960's and the 1980's, so that adjustment of global tropospheric temperature for the ENSO influence (Figure 3) has also resulted in a smaller increase in tropospheric temperature between these two decades, as noted earlier. What is not obvious is whether this increase in SST is due to the greater magnitude and frequency of El Ninos in the 1980's than in the 1960's, or reflects an increase in the background value of SST. If the former is true, and the frequency and magnitude of El Nino is assumed independent of global temperature, then the smaller global temperature increase obtained from the modified data might be considered the more representative from a greenhouse-varning point of view. However, if, as seems more likely, the magnitude and frequency of El Nino varies with global temperature and/or there has been a background increase in SST, then the global temperature change obtained from the unmodified data would be the more representative. In view of the uncertainty in these matters, the modified temperature trace may be of greatest use in the study of shorter-term temperature variations, i.e., in better

resolving volcanic influences on tropospheric temperature and in providing an estimate of the relative warmth of adjacent years independent of ENSO influences.

References

Angell, J.K., Comparison of variations in atmospheric quantities with sea surface temperature variations in the equatorial eastern Pacific, Mon. Wea. Rev., 109, 230242, 1981.

Angell, J.K., Impact of El Nino on the

delineation of tropospheric cooling due to volcanic eruptions, J. Geophys. Res.. 93, 3697-3704, 1988.

Angell, J.K.. Variation in United States

cloudiness and sunshine duration between 1950 and the drought year of 1988, J. Climate, 2. 296-308, 1990.

Angell, J.K., and J. Korshover, Global

temperature variations in the troposphere and stratosphere, 1958-1982, Mon. Wea. Rev., 111, 901-921, 1983.

Folland, C.K., D.E. Parker, and F.E. Kates, Worldwide marine temperature fluctuations 1856-1981, Nature, 310, 670-673, 1984. Hansen, J.E., and S. Lebedeff, Global surface air temperatures: update through 1987, Geophys. Res. Lett. 15, 323-326, 1988. Hoel, P.G., Introduction to Mathematical Statistics, Wiley, New York, 258 pp. 1947. Jones, P.D., The influence of ENSO on global temperatures. Climate Monitor. 17, 80-89, 1989. Jones, P.D., T.M.L. Wigley, C.K. Folland, D.E. Parker, J.K. Angell, S. Lebedeff, and J.E. Hansen, Evidence for global warming in the past decade, Nature, 332, 790, 1988. Navato, A.R., R.E. Newell, J.C. Hsiung, C.B. Billing, and B.C. Weare, Tropospheric mean temperature and its relationship to the oceans and atmospheric aerosols. Mon. Wea. Rev., 109, 244-254, 1981.

Newell, R.E., and B.C. Weare, Factors governing tropospheric mean temperature, Science, 194. 1413-1414, 1976.

Pan, Y.H., and A.H. Oort, Global climate variations connected with sea surface temperature anomalies in the eastern equatorial Pacific for the 1958-73 period, Mon. Wea. Rev., 111, 1244-1258, 1983. Starr, V.P., and A.H. Oort, Five-year (1958-1963 climatic trend for the northern hemisphere, Nature. 242. 310-313, 1973.

Trenberth, K.E., G. Branstator, and P.A. Arkin, Origins of the 1988 North American drought, Science, 242, 1640-1645, 1988.

J. Angell, Air Resources Laboratory, ERL, National Oceanic and Atmospheric Administration, 1325 East West Highway, Silver Spring, MD

20910.

(Received February 13, 1990;

revised June 4, 1990;

accepted June 5, 1990)