References Barton, J. S. (1984) Observing mountain weather using

an automatic weather station. Weather, 39, pp. 140- 145

Dybeck, M. W. and Green, F. H. W. (1955) The Caimgorms weather survey, 1953. Weathc?r, 10, pp. 4148

Harding, R. J. (1978) The variation of the altitudinal gradient of temperature within the British Isles. Gmgr. Ann. Ser. A , 60, pp. 43-49

~ (1979) Radiation in the British uplands. J. Appl.

Harrison, S. J. (1974) Problems in the measurement and evaluation of the climatic resources of upland Britain. In: Taylor, J. A. (Ed.) Climatic resourns and economic acrivicy, David and Charles, pp. 47-63 - (1988) Numerical assessment of local shelter

around weather stations. Weather, 43, pp. 325-330 ~ (1993a) Differences in the duration of snow

Ecol., 16, pp. 161-170

cover on Scottish ski slopes between mild and cold winters. Swtt. Geogr. Mag., 109, pp. 3744 - (1993b) 7he climate of Cennal Region. University

of Stirling Report s~cs/06/93,24 pp. Harrison, S. J. and Harrison, D. J. (1988) The effect of

elevation on the climatically determined growing season in the Ochil Hills. Scoa. Geogr. Mag., 104, pp. 108-115

Johnson, R. C. (1985) Mountain and glen contrasts at Balquhidder. J. Meteoml. (UK) , 10, pp. 105-108

Jones, R. J. A., Tinsley, J. and Court, M. N. (1979) Mesocliiatic studies in the upper Dee basin Aber- deenshire. Meteml. Mag., 108, pp. 289-308

Scottish Office Agriculture and Fisheries Department (1993) Scotland’s Agriculture and Envinmment: Imple- mentation of Agri~Enviironment Regulation (EC) 2078192 in Swhnd, Scottish Office

Tabony, R C. (1985) Relations benveen minimum temperature and topography in Great Britain. J. Climatol., 5, pp. 503-520

Climatic warming in the central Antarctic Peninsula area

P. Stark British Antarctic Survey, Cambridge

One of the earliest research stations established by Britain in the Antarctic was Base F on the Argentine Islands, now known as Faraday (see Fig. 1). Continuous occupation has occurred since 1947 when buildings were erected on the site of the British Graham Land Expedition hut. In 1954 a larger station was established some 700m to the north on a neighbouring island, enabling it to be developed into a size- able geophysical observatory. Throughout this period the instruments were always housed in standard Stevenson screens set on low, rocky outcrops which enabled the wind to scour snow accumulations. Normal observing practices were followed and standard instruments and calibration tests were used. There is no reason to suspect the overall accuracy of the records or that the trends observed are limited to micro- climatological changes within the island group.

The Argentine Islands are located approx- imately 6km west of the Antarctic Peninsula at 65’15’S, 64’16’W. The islands are very small and low-lying, the highest point being 65m above sea-level. The larger ones support per- manent ice domes. In contrast, the peninsula is very mountainous, rising rapidly from the sea to a height of approximately 2000m.

The Intergovernmental Panel on Climate Change (IPCC 1990) compared several general circulation models and their predictions of at- mospheric warming resulting from increased atmospheric carbon dioxide concentrations. These models differed considerably with re- spect to the average global surface temperature increase, but most predicted mean winter tem- perature rises of the order of lOdegC around the periphery of the Antarctic. Other results using coupled atmosphereocean models

215

Fig. 1 tions of the three stations mentioned in the text

The Antarctic Peninsula region, showing bca-

(Stouffer et al. 1989) do, however, suggest that short-term warming around the Antarctic will be somewhat less. Data from stations like Fara- day will be crucial in determining whether this warming is occurring. Synoptic meteorological observations have been made at Faraday since 1947, resulting in one of the longest continuous datasets from the Antarctic. This is of consi- derable significance when studying the cli- matology of the peninsula area.

The warming trend

The dataset used consists of monthly mean air temperatures for Faraday from 1947 to 1990. It is held by the British Antarctic Survey, and is similar to the set by Jones and Limbert (1989), but excludes their modified data from nearby Port Lockroy for 1944-46 and includes values for 1987-90 from Faraday.

For the 44 years of data, the annual mean temperatures range from -1.15"C in 1989 to -8.11"C in 1959 - very large variations for a maritime location. This is mainly due to varia- tions in the extent of sea-ice. In the summer months, when the islands are usually sur- rounded by open sea, variations are small, for example January mean temperatures range

from +2.3 to -1.4"C. Winter sea-ice distribu- tion is much more variable, the effect of which on the monthly mean temperatures is shown, for example, in the July values, which range from -2.6 to -2O.l"C. In effect, for winters with very extensive sea-ice, Faraday becomes more of a continental station.

A least-squares linear regression has been performed on the annual mean temperature values. Despite the large variations in tempera- ture, the analysis gave a line that is statistically a very good fit. The results are:

Gradient = +0.0606degC per year Standard error = +0.017ldegC per year Residual standard deviation = 1.442 degC Statistical significance (from F-test) = 99.90 per cent

The annual mean temperature data, the least- squares regression and standard deviation lines for the data are shown in Fig. 2. The gradient implies that over the 44 years of temperature records the average annual mean temperature has risen by 2.67degC.

Previous analyses using the data from Fara- day and other bases have not revealed any significant trends. In general these analyses have only used data from 1957 onwards. This was the International Geophysical Year (IGY) when many research stations were established in Antarctica. Raper et al. (1984) used data for 1957-82, whilst Sansom (1989) considered the period 1957-86. For these periods analysed, Fig. 2 shows that from 1960 to 1969 the annual temperatures were close to the least-squares regression line, whilst for 1970-75 the values were all above this line and for 1976-82 they were all below. The result of this structure in the temperature records would markedly affect the significance of the results obtained. Also, these analyses were mainly concerned with general Antarctic climatic changes. West coast peninsula stations have a very different climate from stations elsewhere on the continent, with the peninsula mountains exerting a major influ- ence on atmospheric circulation. With such considerable variation in the annual mean tem- peratures, as long a record as possible needs to be analysed.

Although a 44-year dataset is comparatively short by European standards, the annual data

216

. . ----. J 1950 IWO 1970 IWO 1990

YEAR

-a 1940

Fig. 2 Mean annual temperatures at Faraday. The heavy solid line is a least-squares regression line, the broken lines indicate +O. 5 and +1 standard deviations from this line.

were analysed to see if significant periodicity was present. For this, the annual temperature values were converted to differences from the least-squares regression fit to obtain a de- trended time-series. The series was then Fourier analysed with the periodogram ob- tained being shown in Fig. 3.

The broad peak with its maximum at a 22- year return period (2 cycles) may well be the result of an insufficiently long dataset. The two other major peaks have much shorter return periods, approximately 5 and 9 years, although together they still require 45-50 years of data for good analysis. This could have caused some of the difficulties in the analyses performed by Raper et al. (1984) and Sansom (1989), who both used shorter datasets, and even the pre- sent dataset may be too short to calculate the underlying trend accurately.

Seasonal trends

The monthly mean data values were used to calculate decadal means of monthly mean tem- peratures for the decades shown (Table 1). Also shown are the 44-year monthly mean values, and the difference between the monthly values for the first and last decades.

Decadal analysis using data with major Fourier components having periods of 5.548.8

and 22 years is prone to biasing. Calculating other decade sets did show considerable varia- tions in the mean temperature values, but analysis of the resulting graphs led to the same conclusions for the definition of the seasons, as the overall trends were similar. From Table 1 the seasons at Faraday were defined as follows:

Summer (January-March). A fairly steady warming trend over the 40 years of approx- imately 1-1.5degC in total.

1 0 ffl

PERIOD N YEARS

Fig. 3 Periodogram of the Faraday &trended annual mean temperatures

217

Table 1 Monthly values of the decadal means and the 44-year mean Mean decadal temperatures (“C) Difference 44-year mean

(“0

Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. Ann&

1951-60 (1) +0.09 -0.25 - 1.42 -4.88 -7.68

-10.09 -12.86 -10.78

-8.56 -4.88 -2.41 -0.75

-5.37

1961-70 1971-80 1981-90 (2) +0.31 +0.02 -1.10 -2.94 -4.82 -7.37 -9.13

-11.63 -7.75 -5.61 -2.35 -0.19

-4.38

+0.73 +0.80 -0.43 -2.12 -4.52 -6.55

-10.39 -9.88 -7.47 -4.63 -2.37 -0.25

-3.92

+1.14 + 1.08 +0.12 - 1.97 -3.58 -5.11 -7.50 -7.93 -7.31 -4.69 -2.08 -0.08

-3.16

+1.05 +1.33 + 1.54 +2.91 +4.10 +4.98 +5.36 +2.85 +1.25 +0.19 +0.33 +0.67

+2.21

1947-90 +0.55 +0.33 -0.76 -3.10 -5.39 -7.50 -9.97

-10.36 -7.99 -4.97 -2.45 -0.33

-4.33

Autumn (April-June). As the season pro- gresses, the overall monthly temperature warm- ing for the 40 years increases from 3 to 5 degC, but interdecadal variability also increases.

Winter (July and August). An overall warm- ing of between 3 and 6degC, but substantial interdecadal variability exists.

Spring (September-December). Mean monthly temperatures increase rapidly as the season progresses. The interdecadal temperature vari- ations have large relative values, with the warm- ing trend over the 40 years being relatively small.

These season definitions for Faraday cannot be considered representative of the Antarctic, and are probably of limited use for the penin- sula. Considering the peninsula in general, the eastern (Weddell Sea) side has a very different climate. Even considering the western side, the southern end is far more influenced by sea-ice than the Faraday vicinity, whilst the northern end is less affected by sea-ice but is influenced by the Weddell Sea climate. Others (Schwerdtfeger 1970; Raper et al. 1984) have also suggested that the standard seasons are not representative of the Antarctic climate.

Using these season definitions, the results shown in Table 2 were obtained using linear regression analysis on the mean seasonal temperatures.

All the seasons show statistically significant warming at the 95 per cent confidence level, whilst two have values significant at greater than 99.9 per cent. The annual values obtained earlier have been added for completeness.

Comparing the winter and summer seasons, the above data show that the warming trend in winter is some 2.5 times that of summer. Over the 44 years of data this corresponds to the mean winter temperature having risen by 4.10degC whereas the increase for summer is 1.64degC. (The decadal temperature values in Table 1 show this seasonal warming dif- ference.) It is worthwhile noting that the mod- els used in the IPCC (1990) report indicate that the greatest warming trend due to an increase in carbon dioxide would be in the winter, although data from one station cannot be used to confirm this particular finding.

The variations in the statistical significance values with season, as well as the variations in the decadal temperature values, are probably associated with the usual distribution of the sea-ice (King 1994). The west coast of the peninsula is affected by three sources of sea-ice. Firstly, there is the locally produced ice which, if conditions are favourable, can form along the whole length of the peninsula, although its westward extent in the northern regions is unlikely to be great. Secondly, the extreme

Table 2 Linear regression results on 44 years of seasonal data for Faraday Season Gradient Statistical

(degC per significance year) (F-test) (%)

Summer (J, F, M) +0.0372 99.96 Autumn (A, M, J) +0.1053 99.96 Winter (J, A) +0.093 1 96.04 Spring (S, 0, N, D) +0.0284 95.13 Annual +0.0606 99.90

218

north is affected by incursions of Weddell Sea sea-ice whilst, thirdly, the southern areas regu- larly have sea-ice being forced in by winds and currents from the Bellingshausen Sea. Unfor- tunately, good regional sea-ice extent data only exist from 1973 onwards when satellite mea- surements became available. This dataset is still too short to show conclusively a link between the sea-ice extent and the temperature, and it is not possible to say if there are any trends in the cover.

Faraday’s mid-peninsula location, the inter- vening islands, and the ocean currents’ direc- tion mean that Weddell Sea sea-ice is unlikely to reach the area. For most years, and certainly recently, significant sea-ice concentrations from the Bellingshausen Sea to the south rarely arrive and stay before midwinter (21 June). However, it is usual for this source of ice and that locally produced to stay in significant concentrations in the vicinity until mid-Decem- ber. In summer the sea is generally pack-ice free, whilst in autumn any ice formation is generally from the surrounding local area. For the summer and autumn seasons the local climate is thus less affected by outside influ- ences, which may account for the greater statis- tical significance. It is likely that the sea-ice arrival causes the lowering of the statistical sig- nificance and the greater variability in monthly and seasonal temperature values. Raper et al. (1984) and Sansom (1989) found that mainland Antarctic temperatures do not show any statis- tically significant warming trends.

It is worthwhile stating that other datasets held by the British Antarctic Survey for re- search stations at Signy and in the Marguerite Bay area (see Fig. 1 for locations) were also analysed for annual and seasonal trends. These two datasets were similar to updated versions of the sets in Jones and Limbert (1989). The Marguerite Bay dataset is compiled from several stations in the area and is not contin- uous. Both datasets were statistically analysed, but neither gave results at the 95 per cent significance level; the Marguerite Bay value was approximately 93 per cent with an annual increase in temperature of 0.0513degCY whilst that for Signy was 77 per cent with an annual temperature increase of 0.0168degC. Un- doubtedly, the statistical significance of the

data from Marguerite Bay is reduced by gaps in the dataset. A comparison of the annual trends between that of Faraday and Marguerite Bay showed a close matching, although in years with low mean temperatures the Marguerite Bay values were usually substantially lower than those from Faraday, a sign of a greater continental influence. No such matching ex- isted between the Faraday and Signy data. It is clear that this warming trend does include the Marguerite Bay area on the western side of the peninsula, but does not extend across the Wed- dell Sea to Signy.

The climate at Faraday is usually described as maritime Antarctic. This can be seen to be the case in summer with small diurnal and interannual temperature variations, with the average temperatures being similar to the sur- rounding ocean. The winter climate can be seen to be one falling between this maritime model and one which has strong continental influences.

Using Fig. 2 it is possible to define years as normal, warm or cold. For this, I have defined a normal year as one where the annual tem- perature falls between fO. 5 standard deviations of the linear regression. Warm and cold years were those which fall on their respective sides of this. For each year the month with the lowest mean temperature was found. Then, for each temperature category, the coldest month’s month numbers (e.g. July = 7) were summed, and the means and standard deviations calcu- lated. The results shown in Table 3 give a rough estimate of the most likely time for the lowest temperature.

Although this is a crude estimate, the results show that, on average, warm years have their coldest period almost a month later than cold years. This links into the climate having mari- time and continental influences where, on aver- age, a continental climate has its coldest part of the year somewhat earlier than a maritime one.

Table 3 Mean time of occurrence of the coldest month at Faraday ~~

Twe of year Mean month number Sample size Warm 8.0 * 1.1 14 Normal 7.7 * 0.6 18 Cold 7.2 * 0.7 12

219

The scatter is, however, considerable as the standard deviations show. For example, in the warm category six years are coldest in Septem- ber whilst two are coldest in June.

Conclusion

The analysis of the 44-year continuous tem- perature record from Faraday shows a statis- tically significant warming trend over this period of almost 2.7degC. Decadal analysis of the data has shown that the greatest warming is in the autumn and winter months, with values up to 5degC, although this is associated with increased variability and a somewhat reduced statistical significance. This is probably caused by variations in the amount of sea-ice cover and associated weather, the results of which can be seen in the data for the month that is usually the coldest. In the summer and early autumn, when the sea-ice is at a minimum, these conti- nental influences are reduced and the statistical significance of the temperature trend is very high. It is likely that the Antarctic Peninsula plays an important part in reducing the conti- nental influences on the climate; stations fur- ther to the north are affected by weather and ice conditions associated with the Weddell Sea. The analysis of data from Signy shows no statistically significant trends. Data from sta- tions in the Marguerite Bay area show warming trends of a similar magnitude to those observed at Faraday although the statistical significance of these trends is considerably lower, probably due to the broken data record. The trend seen is much greater than that observed in global or Southern Hemisphere temperatures over the same period. It suggests that this is an area of high climatic sensitivity. Climate models also suggest that the marginal sea-ice zone is a region where there may be a large response to a general global warming. It appears that stations in the mid Antarctic Peninsula area are uniquely well located to observe this warming trend.

Acknowledgements

I wish to thank John King (British Antarctic Survey) for his great help in providing the data

and computer analysis in addition to many helpful comments on the paper.

References IPCC (1990) Climate change. The I K C sCient& assess-

ment. Cambridge University Press, 365 pp. Jones, P. D. and Limbert, D. W. S. (1989) Antarctic

sutjiie temperature and pressure data. Oak Ridge Na- tional Laboratory, Environmental Sciences Division, Publication Number 3215

King, J. C. (1994) Recent climate variability in the vicinity of the Antarctic Peninsula. Znt. J. CZimatoZ. (in Press)

Raper, S. C. B., Wigley, T. M. L., Mayes, P. R., Jones, P. D. and Salinger, M. J. (1984) Variations in surface air temperatures: Part 3. The Antarctic 1957-1982. Mon. Wea. Rev., 112, pp. 1341-1353

Sansom, J. (1989) Antarctic surface temperature time series. J. CZimate, 2, pp. 1164-1172

Schwerdtfeger, W. (1970) The climate of the Antarctic. Chutes of the polar regions. World Sum. Climatol., 14,

Stoder, R. J., Manabe, S. and Bryan, K. (1989) Interhemispheric asymmetry in climate response to a gradual increase in atmospheric CO,. Nature, 342, pp. 660-662

pp. 253-355

Royal Meteorological Society Frisby-Green Prize - 1994 arrangements

Since 1983 the Association of British Climatolo- gists has awarded an annual prize for the best undergraduate dissertation in the field of cli- matology. The first F. H. W. Green Prize was awarded in 1984 and this year will be the eleventh such award.

The prize was instituted as a tribute to the wide-ranging climatological work of F. H. W. Green. In 1993 the prize was renamed the Frisby- Green Prize in recognition of the support for the prize and the climatological contributions of Emily Frisby, one of the Association's founder members.

Heads of departments in universities and col- leges in the UK are invited to submit one under- graduate dissertation in the field of climatology of sufficient merit to warrant consideration for this prize.

Further details and an application form can be obtained from: Dr J. McClatchey, School of Environmental Science, Nene College, Moulton Park, Northampton, "2 ~AL.

220