ENSO RELATIONSHIP TO THE RAINFALL OF SRI LANKA

Transcription

1 INTERNATINAL JURNAL F CLIMATLGY Int. J. Climatol. 18: (1998) ENS RELATINSHIP T THE RAINFALL F SRI LANKA R.P. KANE* Instituto Nacional de Pesquisas Espacias INPE, Caixa Postal 515, São José dos Campos, SP, Brazil Receied 14 June 1997 Reised 31 ctober 1997 Accepted 8 Noember 1997 ABSTRACT Each year during was examined to check whether it had an El Niño (EN) and/or a Southern scillation Index (SI) minimum (S) and/or warm (W) or cold (C) equatorial eastern Pactfic sea surface temperatures SST. Several years were ENSW, which were further subdivided into two groups namely, unambiguous ENSW where El Niño existed and SI minima and SST maxima were in the middle of the calendar year (May August), and ambiguous ENSW where El Niño existed but the SI minima and SST maxima were in the early or late part of the calendar year, not in the middle. ther El Niño events were of the type ENS, ENW, ENC and EN. Some years not having El Niño were of the types SW, SC, S, W and C, the last one (C) containing all anti-el Niños, i.e. La Niñas. Remaining years were termed as non-events. For all these years, the normalized rainfall deviations from the mean were examined for seven rainfall series, three in India and four in Sri Lanka. For the all India summer monsoon rainfall and Indian southern peninsula summer monsoon rainfall, unambiguous ENSW years showed a very good association with droughts. The Sri Lanka southwest monsoon rains also showed a similar tendency; but the Sri Lanka second intermonsoon season rainfall showed a strong opposite tendency (floods instead of droughts). For other types of El Niño years, results were generally obscure. For C type events, results were opposite to those of unambiguous events, as expected. It is suggested that for obtaining composite maps based on El Niño years, only the unambiguous ENSW years may be used Royal Meteorological Society. KEY WRDS: rainfall; Sri Lanka; ENS; Southern scillation; peninsular India 1. INTRDUCTIN The rainfall characteristics in different parts of India are very different. ver major parts of India, the Asian southwest monsoon contributes ca. 75% of the annual rainfall. However, in the southern part of India, ca. 60% rainfall is in the summer (June, July, August, September; JJAS) months, ca. 25% in ctober, November, December and ca. 15% in January May. The rainfall characteristics in Sri Lanka, an island near the southernmost tip of India, have some resemblance with the rainfall characteristics in southern India. Rasmusson and Carpenter (1983) examined the relationship between eastern equatorial Pacific sea surface temperature anomalies (El Niño events) and rainfall over India and Sri Lanka. For India as a whole, they reported a strong tendency for a below normal summer monsoon rainfall during the El Niño years. n the other hand, the anomaly pattern over Sri Lanka and extreme southern India was quite different, the major feature being above normal precipitation during the autumn of El Niño years. Ropelewski and Halpert (1987, 1989) also reported similar results. In a further analysis, Suppiah (1987) showed that except for the east coast region, all other parts of Sri Lanka experienced a bimodal pattern in the annual rainfall cycle coincident with the northward and southward migrations of the ITCZ. The correlation between the March September Sri Lanka rainfall and the Southern scillation Index * Correspondence to: Instituto Nacional de Pesquisas Espacias INPE, Caixa Postal 515, São José dos Campos, SP, Brazil. kane@laser.inpe.br Contract grant sponsor: FNDCT Brazil; Contract grant number: Contract FINEP 537/CT CCC /98/ $ Royal Meteorological Society

2 860 R.P. KANE (SI, represented by the Tahiti minus Darwin atmospheric pressure difference, TD) was, while the correlation between rainfall in the rest of the year and SI was. In general, due to the strong seasonality in Sri Lanka rainfall, correlations between SI and rainfall showed and values in space and time (Suppiah, 1988, 1989). Recently, Suppiah (1996) presented variations in the temporal and spatial relationships between Sri Lanka rainfall and SI during the 110-year period and reported four distinct epochs, namely when Indian and Sri Lanka rainfalls were above normal, when Sri Lanka was above and India was below, when the reverse was true, and when both were below normal. Also, major changes occurred in spatial patterns of rainfall SI correlations. Later, Suppiah (1997) examined the influence of extreme phases of the SI (El Niño and La Niña events) on the seasonal rainfall in Sri Lanka and reported that the southwest monsoon season (JJAS) rainfall had a SI relationship opposite to that of the second intermonsoon season rainfall. Recently, we noticed that not all El Niño years gave the same type of rainfall response (floods or droughts) in areas where ENS relationships are reported. A finer classification of El Niño years could be made where events of the unambiguous ENSW type (described in the next section) gave very good associations with droughts in Indian summer monsoon rainfall, Australian rainfall and rainfall in other parts of the world (Kane, 1997a,b, 1998). In this communication, we examine whether this finer classification yields better SI relationships for the Sri Lanka rainfall. The main purpose of the present paper is an examination of the Sri Lanka rainfall patterns and hence, more attention is paid to the same. All-India summer monsoon rainfall and southern India rainfalls are used for comparison because of the vicinity of these regions. 2. FINER CLASSIFICATIN F ENS EVENTS The term ENS is used nowadays for the general phenomenon of the Walker Circulation. However, in our classification here, its components El Niño, Southern scillation minima are used in their literal sense. Thus, every year was examined to check whether it had an El Niño (as listed in Quinn et al., 1978, 1987), and/or Southern scillation Index minimum (S) and/or warm (W) or cold (C) equatorial eastern Pacific sea surface temperatures (SST). An examination of the monthly means often showed erratic behaviour. Hence, 12-month running means were calculated and used to locate the maxima or minima. Several years had ENSW, i.e. El Niño (EN) existed and SI had minima (S) and Pacific SST were warm (W). These were subdivided into two groups namely, unambiguous ENSW where El Niño existed and the SI minima and SST maxima were in the middle of the calendar year (May August) and, ambiguous ENSW where El Niño existed, but the SI minima and SST maxima were in the early or later part of the calendar year, not in the middle. Besides these, there were other El Niño years of the type ENS (i.e. SI minima existed but SST was neither warm nor cold, just normal), ENW, ENC (i.e. SI minima did not exist but SST was warm W or cold C) or just EN (i.e. only El Niño existed and there were no SI minima or SST maxima or minima). Some other years did not have an El Niño and were of the types SW, SC, S, W, and C where the last category C contains all anti-el Niños, i.e. La Niñas. Years not falling into any of these categories were termed as non-events. We have this classification ready for all years from 1871 onwards, though only 1881 onwards is used in this communication. Figure 1 illustrates the procedure of classification. The four panels 1, 2, 3, 4 refer to , , and In each panel, the top curve is for the Southern scillation Index (SI) obtained by Wright (1977) from a principal component analysis (PCA) of seasonal mean pressures at eight locations: Cape Town, Bombay, Djakarta, Darwin, Adelaide, Apia, Honolulu and Santiago. This index is available for only. However, Chen (1982) showed that the Tahiti minus Darwin atmospheric pressure difference (TD) was simpler and served equally well. Parker (1983) gave (T D) values for and further values were available from meteorological data reports. Whereas the monthly values of (TD) and the Wright Index did not match very well, the 12-monthly running means of (TD) matched very well with the Wright Index. In panel 4, the values for 1974 onwards are the 12-monthly running means of (TD). The next two plots show the SST index of Wright (1984) for the eastern equatorial Pacific, and a similar index given by Angell (Angell, 1981, and further private

3 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 861 Figure 1. Plots of 12-monthly running means of Southern scillation Index (SI) and sea surface temperatures in equatorial eastern Pacific (SST) for (panels 1 4) and the characterization of each year as having El Niño (EN) and/or SI minima (S) and/or warm (W) or cold (C) Pacific SST (rectangles). For years having El Niño, the symbols above the rectangles indicate strength (S, strong; M, moderate; W, weak) of the El Niño. SI minima and SST maxima are painted black, and SST minima are shown hatched

4 862 R.P. KANE communication). El Niño years were obtained from Quinn et al. (1978, 1987). The rectangles at the bottom of each panel indicate the classification for each year, e.g was an El Niño year (EN). It also had a SI minimum (S) and a SST maximum (W), both shown black. Hence, 1957 was an ENSW. Also, the S and W occurred in the middle of the year. So, 1957 is an unambiguous ENSW. The next year 1958 was also an El Niño year and had S and W and therefore was an ENSW. But the S and W occurred in the early part of the year, not in the middle. Hence, 1958 was an ambiguous ENSW. The year 1889 had cold SST (i.e. C) and is used by Suppiah (1997) as a La Niña event. But, according to Quinn et al. there was an El Niño in that year. Therefore, 1889 is an ENC. In 1944, there was a SI minimum (S) and a SST maximum (W); since there was no El Niño in that year, 1944 was a SW. In 1986, there was no El Niño and SI values began to drop only near the end, but SST had already developed. So, 1996 was only a W. In 1987, there was an El Niño, SI minima and SST maxima occurred almost in the middle of the year and hence, 1987 was an unambiguous ENSW. These classifications are somewhat subjective and some events were on the borderline of two categories. In general, the classification gave very good results for the Indian rainfall and is thus, fully justified. In Figure 1, wherever there is an El Niño, the symbols S (strong), M (moderate), W (weak) on the top of the rectangle indicate the strength of the El Niño involved. 3. RAINFALL DATA Rainfall data for All-India (AI) and for the southern Indian Peninsula (SP) were obtained from Singh and Sontakke (1996) and Sontakke and Singh (1996), and data for Sri Lanka rainfall were read out from Suppiah (1997). All these data are one value per year of the normalised rainfalls, i.e. departures from mean of the series, expressed in units of standard deviation of the series. For Sri Lanka rainfall, four series are used, corresponding to the months MAM (March, April, May, termed as first intermonsoon FIM by Suppiah); JJAS (June, July, August, September, termed as southwest monsoon SWM by Suppiah); N (ctober, November, termed as second intermonsoon by Suppiah) and DJF (December, and January February of next year, termed as northeast monsoon NEM by Suppiah). These series were obtained as the average of 29 stations in Sri Lanka, selected after comparing the space and time coefficients of a principle component analysis. Details of the data sources and definition of seasons are given in Suppiah (1996). It is assumed that the average is a good representation of a homogeneous region. In India, southern India is a fairly coherent region and all India monsoon rainfall also contains a large homogeneous part. Details of the rainfall characteristics in 29 meteorological subdivisions of India and their El Niño relationships are discussed in Kane (1998). 4. RAINFALL DEVIATINS 4.1. Eents inoling El Niños Table I shows the nature of the deviations of rainfall in the seven series, namely, All-India summer monsoon (AI-JJAS), Indian southern peninsula summer monsoon (SP-JJAS) and winter rainfall (SP- ND), and Sri Lanka rainfalls (L-MAM, L-JJAS, L-N, L-DJF) for the various types of years (ENSW etc.). Positive and deviations within 0 to are designated as and, while and deviations exceeding are designated as (floods) and (droughts). The following may be noted. (i) Table I(a) refers to years of unambiguous ENSW. There were 16 events of the unambiguous ENSW type. In some cases, two successive years are El Niño years ( , etc.). For these double events, the first year is indicated as I and the second as II. The symbol R indicates that these events were selected as El Niño years by Suppiah (1997) also. Interestingly, 15 events are in the Suppiah list too. ur 16th event is 1987, chosen from the list of Quinn et al. (1987). Instead, Suppiah selected 1986 as an El Niño year; but for us it is only a W, included in a later table.

5 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 863 Table I. Rainfall status during El Niño years of the types (a) unambiguous ENSW; (b) ambiguous ENSW; and (c) other types of El Niños. S (strong), M (moderate), W (weak) indicate the strengths of the El Niños involved. I and II indicate first and second years of double events (El Niños in two consecutive years). R indicates that this event was selected by Suppiah (1997). Columns represent rainfalls of (1) All-India summer monsoon (AI-JJAS); (2) and (3) Indian South Peninsular (SP-JJAS) and (SP-ND); (4) (5) (6) and (7) Sri Lanka first intermonsoon (L-MAM), southwest monsoon (L-JJAS), second intermonsoon (L-N) and northeast monsoon (L-DJF), respectively. The symbols and indicate and deviations within 0 and, while () and () represent deviations exceeding 1 (AI-JJAS) 2 (SP-JJAS) 3 (SP-ND) 4 (L-MAM) 5 (L-JJAS) 6 (L-N) 7 (L-DJF) (a) Unambiguous ENSW S 1888 R M 1896 R S 1899 R M 1902 R M 1905 R S 1911 R S 1918 I R M 1930 I R S 1941 II R M 1951 R S 1957 I R M 1965 R S 1972 I R M 1976 R S 1982 I R M events Deviations Deviations Positive/total (b) Ambiguous ENSW M 1914 M 1919 II M 1923 R S 1925 I R S 1926 II M 1931 II S 1940 I W 1948 M 1953 R S 1958 II W 1963 R W 1969 R S 1983 II 13 events Deviations Deviations Positive/total

6 864 R.P. KANE Table I. (Continued) 1 (AI-JJAS) 2 (SP-JJAS) 3 (SP-ND) 4 (L-MAM) 5 (L-JJAS) 6 (L-N) 7 (L-DJF) (c) ther El Niños S 1891 ENS R S 1900 ENS S 1912 ENS S 1884 ENW R M 1897 EN S 1932 EN R M 1939 EN R M 1943 EN Eight events Deviations Deviations Positive/total M 1887 ENC M 1889 ENC M 1907 ENC S 1917 ENC S 1973 ENC Five events Deviations Deviations Positive/total (ii) For the All-India summer monsoon (AI-JJAS), 15 events show deviations (1, 15), nine of these being severe droughts (circles). Thus, there is a very strong association between droughts in All-India summer monsoon and El Niños of the unambiguous ENSW type. The same is roughly true for the Indian southern peninsula summer monsoon (SP-JJAS) where 13 events showed deviations (3, 13). But the winter rainfall (SP-ND) showed (9, 7), indicating poor relationship. In the Sri Lanka rainfall, L-MAM months showed (9, 7) and L-DJF months showed (6, 10), rather poor relationships. But L-JJAS showed (4, 12), indicating a bias for droughts, same as for AI-JJAS and SP-JJAS. Instead, L-N showed (14, 2), an overwhelming bias for excess rains. Thus, during unambiguous ENSW, the whole of India (including the southern peninsula) and Sri Lanka show a tendency for droughts during the southwest monsoon (JJAS months). For later months, southern peninsula India has substantial rainfalls but poorly related to such events, while Sri Lanka rainfalls show a tendency for excess rains, opposite to that for the JJAS months. The monthly composite rainfall anomalies for El Niño events obtained by Suppiah (1997) also showed this pattern, though the pattern is seen more strongly here. The bottom part of Table I(a) shows the statistics, namely numbers of and deviations and the fraction of s (e.g. for AI-JJAS, 1, 15, fraction of s 1/16=0.06). These are discussed later. (iii) Table I(b) shows results for 13 ambiguous ENSW events. Here, and deviations occur roughly equally, indicating a poor relationship. In particular, five events here are the second years (II) of double events and, at least for the AI-JJAS, show deviations, opposite to those for AI-JJAS in Table I(a). Thus, these second years have a tendency not to show droughts and may even show floods (e.g II). For the Sri Lanka rainfall also, these second years showed deviations for L-N, opposite to the deviations for L-N in Table I(a). This is similar to the

7 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 865 tendency towards opposite anomalies between the current and previous year, particularly from ctober to December mentioned by Suppiah (1997). In general, ambiguous ENSW do not show a good relationship with rainfall. However, L-DJF seems to be an exception, showing (10, 3), almost opposite to (6, 10) of Table I(a). The two combined (ab) would show (16, 13), a poor relationship on the whole. Suppiah (1997) has chosen five events (R) in this group and four of these seem to show deviations for L-N, similar to those in Table I(a). (iv) Table I(c) upper part, shows results for eight other types of El Niño years namely, ENS, ENW, and EN. Here, SP-ND (6, 2) and L-N (7, 1) show bias for excess rains, the latter similar to Table I(a). verall, for the 37 events in Table I(a,b,c), AI-JJAS shows (14, 23), SP-JJAS (15, 22), SP-ND (22, 15), L-MAM (20, 20), L-JJAS (13, 24), L-N (27, 10) and L-DJF (22, 15). Thus, the only striking relationship is of L-N. But AI-JJAS and L-N are really striking in Table I(a) only, in opposite ways (droughts and excess rains, respectively). (v) The lower part of Table I(c) shows five events of the type ENC, i.e. an El Niño occurred but there was no accompanying SI minimum, and the SST was cold (C). These two are contradictory effects. In case of AI-JJAS, excess rains predominated (4, 1) indicating that the effect of cold SST prevailed and El Niño proved ineffective. In case of L-N also (1, 4), the effect of cold SST prevailed, as here, El Niño was supposed to give excess rains as in Table I(a). (vi) In general, the MAM and DJF rainfalls of Sri Lanka (FIM and NEM seasons of Suppiah) seem to be poorly related to any type of El Niño events. Suppiah (1997) also came to a similar conclusion Eents not inoling El Niños (S, W, C) Table II (upper part) shows results for 12 years when El Niños did not occur but SI did show minima (S) and/or SST was warm (W). There were six SW, three W and three S. All these are expected to show results similar to El Niños. However, no striking biases are seen in this group as a whole, though perhaps L-N (8, 4) and L-DJF (3, 9) may be considered similar to Table I(a). The lower part of Table II shows results for four SC type years and the results are expected to be contradictory for the SI minima and cold SST. For AI-JJAS, the effect of cold SST seems to have prevailed. For L-JJAS, the effect is similar to that of Table I(a) and hence, the SI minima effect must have prevailed. Interestingly, L-DJF shows excess rains; but we are not sure what to expect here Non-eents As mentioned earlier, several years have neither El Niño nor a Southern scillation minimum nor warm SST nor cold SST. These are non-events and should not show rainfall extremes and/or should not show any biases for large or deviations. Table III shows the results for 21 non-events. As expected, the and deviations are seen almost equally in all series, the maximum difference being (8, 13). However, the deviations are not always small, indicating that severe droughts or floods can occur due to other causes. In the Indian region, five or less rainfall extremes occurred; but in Sri Lanka, the number reached 12 (five floods, seven droughts) for L-DJF. Thus, the rainfall of the island of Sri Lanka is affected considerably by other (probably local) factors Cold eents (La Niña) Table IV shows the results for 32 years exclusively of the cold SST type. Suppiah (1997) used 22 La Niña events in his analysis, of which 17 are included in Table IV (marked as R). His La Niña events of 1892 and 1973 are our ENC events, his 1920 is our W, his 1931 is our ambiguous ENSW and his 1949 is our SC. For AI-JJAS, the score is (30, 2), an overwhelming association with excess rains. For SP-JJAS, the effect is less striking (20, 12) and the relationship is poor (13, 19) for SP-ND and L-MAM and for L-JJAS (16, 16). However, for L-N (8, 24), the effect is strikingly biased towards droughts, the opposite of AI-JJAS. For L-DJF also (11, 21), there is a bias for droughts. Thus, the La Niña effects are, in general, opposite to those for El Niños and, AI-JJAS and L-N show effects opposite to each other.

8 866 R.P. KANE In the tables, the number of and deviations and the fraction of s is given at the bottom. Figure 2 shows these graphically for the seven series for the various categories of years. The fractions are from 0 to The central vertical line indicates a fraction 0.50, i.e. equal number (50%) of and rainfall deviations, implying very poor relationship. Fractions far away from this line to the left imply small number of deviations, i.e. a large number of deviations. Far away to the right implies a small number of deviations, i.e. a large number of deviations. Thus, in Figure 2(a) for unambiguous ENSW, All-India summer monsoon rainfall (AI-JJAS, marked by a rectangle) and, to a lesser extent SP-JJAS and L-JJAS, show preponderance of deviations (rainfall deficit). In contrast, L-N (marked by a circle) shows a preponderance of deviations (rainfall excess). In Figure 2(b) for ambiguous ENSW and Figure 2(c) for other types of El Niños (ENS, ENW, EN), the fractions are near the 0.50 line, indicating poor relationship, though L-N and to a lesser extent L-DJF show a bias for higher fractions, as in Figure 2(a). In Figure 2(d) for ENC, the situation is almost reverse, indicating the influence of cold SST rather than El Niño. In Figure 2(e) for SW, S, W which are expected to give results similar to El Niño, relationship is poor, except for AI-JJAS which shows results similar to Figure 2(a). In Figure 2(f) for SC, AI-JJAS shows results opposite to those of Figure 2(a), implying effect of cold SST rather than the SI minima, while L-JJAS shows results similar to those of Figure 2(a), implying effect of the SI minima rather than cold SST. In Figure 2(g) for non-eents, fractions are cluttered near 0.50, as expected. In Figure 2(h) for cold SST (i.e. La Niña events), AI-JJAS has large fractions of deviations and L-N has small fractions of deviations, both opposite to those in Figure 2(a), as expected. The L-MAM has, in general, a Table II. As in Table I, for years not having an El Niño namely, SW, W, S and SC 1 (AI-JJAS) 2 (SP-JJAS) 3 (SP-ND) 4 (L-MAM) 5 (L-JJAS) 6 (L-N) 7 (L-DJF) 1888 SW 1904 SW 1913 SW 1944 SW 1977 SW 1979 SW 1920 W 1968 W 1986 W 1885 S 1959 S 1974 S 12 events Deviations Deviations Positive/ total SC 1936 SC 1946 SC 1949 SC Four events Deviations Deviations Positive/ total

9 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 867 Table III. As in Table I, for non-events (no El Niño,noS,noWorC) 1 (AI-JJAS) 2 (SP-JJAS) 3 (SP-ND) 4 (L-MAM) 5 (L-JJAS) 6 (L-N) 7 (L-DJF) events Deviations Deviations Positive/ total poor relationship; but L-DJF does show some strong relationships. The behaviour of individual series is given in Table V, where fractions are considered as indicative of no clear relationship (x), preponderance of deviations (fractions of s less than 0.40) is indicated as droughts (D) and preponderance of deviations (fractions of s exceeding 0.60) is indicated as floods (F). In Figure 2, the distributions are, in general, broad, except for the non-events, where fractions are concentrated near 0.5. A 2 -test was applied and the distributions for the unambiguous ENSW (Figure 2(a)) and the cold SST events (Figure 2(h)) were found to be significantly different from that of non-events (Figure 2(g)) at 99% and 95% confidence levels, respectively. Some others were different at a 90% level. 5. CNCLUSINS AND DISCUSSIN Each year during was checked to see whether it had an El Niño (EN) and/or a Southern scillation Index minimum (S) and/or warm (W) or cold (C) equatorial eastern Pacific sea surface temperatures. Several years were ENSW, which were subdivided into two groups namely, unambiguous ENSW where El Niño existed and the SI minima and SST maxima were in the middle of the calendar year (May August) and, ambiguous ENSW where El Niño existed, but the SI minima and SST maxima were in the early or later part of the calendar year, not in the middle. Some other El Niño years were of the types ENS, ENW, ENC, EN and some years without an El Niño were of the types SW, SC, S, W, C, the last (C) containing all anti-ei Niños, i.e. La Niñas. ther years were termed as non-events.

10 868 R.P. KANE For rainfall, seven series were considered, All-India summer monsoon rainfall (AI-JJAS), Indian southern peninsula summer monsoon and autumn rainfalls (SP-JJAS and SP-ND) and Sri Lanka first intermonsoon, southwest monsoon, second intermonsoon and northeast monsoon rainfalls (L-MAM, L-JJAS, L-N and L-DJF). Each series was expressed as deviations from the mean, in units of the standard deviation of the series (normalised rainfall). Positive and deviations within 0 were designated as and, and those exceeding were designated as () and (). For events in each category of years, the numbers of and deviations for each series were counted. The following was noticed. (i) For 16 unambiguous ENSW years, the All-India summer monsoon rainfall (AI-JJAS) and Indian southern peninsular summer monsoon rainfall (SP-JJAS) were predominantly in deficit, and the Sri Lanka L-N rainfall was predominantly in excess. Table IV. As in Table I, for cold SST events (La Niñas, colder waters in equatorial eastern Pacific) 1 (AI-JJAS) 2 (SP-JJAS) 3 (SP-ND) 4 (L-MAM) 5 (L-JJAS) 6 (L-N) 7 (L-DJF) R R R 1903 R 1906 R 1908 R R R R 1933 R R 1942 R R R R R 1988 R 32 events Deviations Deviations Positive/ total

11 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 869 Figure 2. Fraction of rainfall deviations in seven rainfall series AI-JJAS, SP-JJAS, SP-ND, L-MAM, L-JJAS, L-N, L-DJF for each category of years. (a) Unambiguous ENSW; (b) ambiguous ENSW; (c) ENS, ENW, EN; (d) ENC, (e) SW, W, S; (f) SC; (g) non-events; and (h) C events (La Niña). The central vertical line indicates the fraction 0.50, implying equal number of and rainfall deviations, and hence, poor relationship. All-India summer monsoon season rainfall (AI-JJAS) is shown by a rectangle and Sri Lanka second intermonsoon season rainfall (L-N) by a circle (ii) For ambiguous ENSW and other types of El Niño years (ENS, ENW, EN), AI-JJAS and SP-JJAS had almost an equal number of and deviations, with slight bias for deviations, especially for second years of double events (El Niño occurring in two consecutive years). L-N had a bias for excess rainfall. (iii) For ENC events, the response was the reverse, indicating the overpowering influence of cold SST rather than El Niño. (iv) For SW, S, W events, responses were similar to those for the unambiguous ENSW while for SC, responses were not clear. (v) For non-events, and deviations were almost equal. (vi) For cold SST events, responses were opposite to those for unambiguous ENSW. Suppiah (1997) examined the response of Sri Lanka rainfall to the extremes of the Southern scillation phenomenon (El Niño and La Niña). He reported that and rainfall anomalies during the

12 870 R.P. KANE Table V. The response of the seven rainfall series (AI-JJAS etc.) to the various categories (ENSW etc.) of years. The symbols D, x, F represent fractions of deviations less than 0.40, 0.40 to 0.60, exceeding 0.60, respectively Unambigous ENSW Ambiguous ENSW ENS, ENW, EN ENC SW, W, S SC Non-events C AI-JJAS D F x F D F x F SP-JJAS D F D x x x D F SP-ND x x F x x D x x L-MAM x x F x x x x x L-JJAS D x D x x D x x L-NF F x F D F x D D L-DJF D F F D D F x D southwest monsoon season (JJAS months) were associated with La Niña and El Niño events; however, for the second intermonsoon season (N months), the response was the opposite. Also, rainfall anomalies during the first intermonsoon (MAM months) and the northeast monsoon (DJF months) showed no clear contrasting patterns. ur results are in general agreement with his results and show, in addition, that the effects are seen much more clearly for the unambiguous ENSW years. Thus, not all El Niño years will necessarily reveal clear-cut rainfall anomaly associations. It seems that the presence of an El Niño (usually in the early part of the calendar year) followed by a Southern scillation Index minimum and equatorial eastern Pacific SST maximum in the middle of the calendar year (MJJA) are necessary for the proper development of the Walker Circulation (Bjerknes, 1969) which causes the rainfall anomalies in some parts of the globe. The composite maps of SST anomalies over the Pacific and Indian ceans for wet and dry years for the Sri Lanka southwest monsoon and second intermonsoon seasons presented by Suppiah (1997) probably indicate the same feature. In view of the fact that, among the various El Niño events, the unambiguous ENSW show the best associations, we suggest that for obtaining composites in studies related to El Niño and La Niña, only the unambiguous ENSW listed in our Table I(a) and the cold SST events listed in Table IV may be used. As mentioned in Suppiah (1997, and references therein), during the active phase of the southwest summer monsoon, central India receives heavy rains while southern India and Sri Lanka receive less rains. During the break phase, this pattern is reversed due to the meridional circulation pattern (e.g. Ananthakrishnan, 1970; Ragahavan, 1973; Sumathipala and de Silva, 1981). However, this intraseasonal oscillation is dominant throughout the summer monsoon season (e.g. Sikka and Gadgil, 1980; Yasunari, 1980) and is modulated by the Walker Circulation (e.g. Parthasarathy and Pant, 1985; Suppiah 1996). Thus, relationships between the ENS phenomenon and rainfalls in India and Sri Lanka during the summer monsoon and autumn are expected. During the first intermonsoon season (MAM months) in Sri Lanka, rainfall is received from weather disturbances that originate within the ITCZ and convective activity and thunderstorms are frequent. Hence, ENS relationships, if any, are obscured. During the northeast monsoon season (DJF months), the ITCZ is located further south of Sri Lanka and northeast monsoon winds bring rainfall to the eastern part. However, tropical cyclones and depressions also bring rainfall in this season and their frequency has a strong interannual variability, obscuring ENS effects, if there are any. However, our results show that the ENS relationship is better for the Sri Lanka northeast monsoon (DJF) than for its first intermonsoon (MAM) for some categories, if the finer classification is adopted. Suppiah (1996) identified four distinct epochs , , and during which the Indian and Sri Lanka rainfalls had different phase relationships and the correlations with SI also changed from one epoch to another. In our analysis, no such epoch dependence is evident. For example, for the Indian summer monsoon (JJAS), 15 out of 16 unambiguous ENSW showed deviations, the only exception being For Sri Lanka rainfall in the second intermonsoon (N), 14 out of 16 unambiguous ENSW showed deviations, the exceptions being 1918 and For Sri Lanka southwest monsoon rainfall (JJAS), 12 out of 16 unambiguous ENSW showed deviations (similar to AI-JJAS) and the four exceptions were 1896, 1902, 1941, If all 42 El

13 ENS RELATINSHIP T THE RAINFALL F SRI LANKA 871 Niño events are considered, All-India (JJAS) and Sri Lanka (JJAS) had similar deviations in 1888, 1889, 1891, 1897, 1899, 1900, 1905, 1911, 1917, 1918, 1919, 1925, 1926, 1930, 1931, 1932, 1939, 1940, 1953, 1957, 1963, 1965, 1969, 1972, 1982, 1987, and dissimilar in 1884, 1887, 1896, 1902, 1907, 1912, 1914, 1923, 1941, 1943, 1948, 1951, 1958, 1973, 1976, Both these seem to be spread over all the epochs and no bias for any particular interval is discernible. The results presented here refer to data up to 1990 only. Since then, there have been El Niño events near 1992 and recently in 1997 (a very strong event). Results for these would be examined as and when data are available. ACKNWLEDGEMENTS This work was partially supported by FNDCT Brazil under Contract FINEP 537/CT. REFERENCES Ananthakrishnan, R Some aspects of the monsoon circulation and monsoon rainfall, Pure Appl. Geophys., 115, Angell, J.K Comparison of variations in atmospheric quantities with sea surface temperature variations in the equatorial eastern Pacific, Mon. Wea. Re., 109, Bjerknes, L Atmospheric teleconnections from the equatorial Pacific, Mon. Wea. Re., 97, Chen, W.Y Assessment of Southern scillation sea level pressure indices, Mon. Wea. Re., 110, Kane, R.P. 1997a. n the relationship of ENS with rainfall over different parts of Australia, Aust. Met. Mag., 46, Kane, R.P. 1997b. Relationship of El Nino Southern scillation and Pacific sea surface temperature with rainfall in various regions of the globe, Mon. Wea. Re., 125, Kane, R.P Extremes of the ENS phenomenon and Indian summer monsoon rainfall, Int. J. Climatol., 18, in press. Parker D.E Documentation of a Southern scillation index, Meteorol. Mag., 112, Parthasarathy, B. and Pant, G.B Seasonal relationships between summer monsoon rainfall and the Southern scillation, J. Climatol., 5, Quinn W.H., Zopf, D.., Short, K.S. and Kuo Yang R.T.W Historical trends and statistics of the Southern scillation, El Nino and Indonesian Droughts, Fish. Bull., 76, Quinn W.H., Neal V.T. and Antunez de Mayolo, S.E El Nino occurrences over the past four and a half centuries, J. Geophys. Res., 92, Ragahavan, K Break-monsoon over India, Mon. Wea. Re., 101, Rasmusson, E.M. and Carpenter, T.H The relationship between eastern equatorial Pacific sea surface temperatures and rainfall over India and Sri Lanka, Mon. Wea. Re., 111, Ropelewski, C.F. and Halpert, M.S Global and regional scale precipitation patterns associated with El Nino/Southern oscillation, Mon. Wea. Re., 115, Ropelewski, C.F. and Halpert, M.S Precipitation patterns associated with the high index phase of Southern scillation, J. Climate, 2, Sikka, D.R. and Gadgil, S n the maximum cloud zone and the ITCZ over Indian longitudes during the southwest monsoon, Mon. Wea. Re., 108, Singh, N. and Sontakke, N.A The instrumental period rainfall series of the Indian region: A documentation, Research Report No. RR-067, Contribution from Indian Institute of Tropical Meteorology, Pune , 79 pp. Sontakke, N.A. and Singh, N Longest instrumental regions and All-India summer monsoon rainfall series using optimum observations: reconstruction and update, Holocene, 6.3, Sumathipala, W.L. and de Silva, M.B.G Rainfall pattern over Sri Lanka during the break period of the Indian summer monsoon, Proceedings, International Conference, Scientific Results of Monsoon Experiment, Denpasar, Bali, Indonesia. Suppiah, R Atmospheric circulation ariations and the rainfall of Sri. Lanka, Sci. Rep. Inst. Geosci. University of Tsukuba, Section A9, Suppiah, R Relationships between Indian cean sea surface temperature and the rainfall of Sri Lanka, J. Meteorol. Soc. Jpn., 66, Suppiah, R Relationships between the southern oscillation and the rainfall of Sri Lanka, Int. J. Climatol., 9, Suppiah, R Spatial and temporal variations in the relationships between the Southern scillation phenomenon and the rainfall of Sri Lanka, Int. J. Climatol., 16, Suppiah, R Extremes of the Southern scillation phenomenon and the rainfall of Sri Lanka, Int. J. Climatol., 17, Wright, P.B The Southern scillation patterns and mechanisms of the teleconnections and the persistence, Report HIG 77-13, Hawaii Institute of Geophysics, 107 p. Wright, P.B Relationship between indices of the Southern scillation, Mon. Wea. Re., 112, Yasunari, T A quasi-stationary appearance of 30- to 40-day period in the cloudiness fluctuations during the summer monsoon over India, J. Meteorol. Soc. Jpn., 59,