arX

iv:p

hysi

cs/9

8120

40v1

[ph

ysic

s.ao

-ph]

21

Dec

199

8

On the El-Nino Teleconnection to

Spring Precipitation in Europe

Geert Jan van Oldenborgh

Gerrit Burgers

Albert Klein Tank

KNMI, De Bilt, Netherlands

December 1998

Abstract

In a statisti al analysis of more than a entury of data we �nd a strong onne tion

between strong warm El Ni~no winter events and high spring pre ipitation in a

band from Southern England eastwards into Asia. This relationship is an exten-

sion of the onne tion mentioned by Kiladis and Diaz (1989), and mu h stronger

than the winter season tele onne tion that has been the subje t of other studies.

Linear orrelation oeÆ ients between DJF NINO3 indi es and MAM pre ipita-

tion are higher than r = 0.3 for individual stations, and as high as r = 0.49 for

an index of pre ipitation anomalies around 50

◦N from 5

◦W to 35

◦E. The lagged

orrelation suggests that south-east Asian surfa e temperature anomalies may

a t as intermediate variables.

1 Introduction

The re ent strong El Ni~no has stirred up again interest in possible tele on-

ne tions to Europe. A few in uen es have been mentioned in previous studies.

During the winter season, S andinavia was found to be older and dryer just

after old (La Ni~na) events (Berlage, 1966), and an in rease in temperature

and pre ipitation with El Ni~no onditions was found in this area (van Loon

and Madden, 1981; Halpert and Ropelewski, 1992). Central European winters

would have the opposite tenden ies: warmer and wetter (Kiladis and Diaz, 1989)

during ENSO warm events; this squares with the in rease of y loni Gro�wet-

1

ter days observed in Germany (Fraedri h, 1994) and England (Wilby, 1993)

during El Ni~no years.

For the spring season after an El Ni~no event (Kiladis and Diaz, 1989) note

pre ipitation anomalies in entral Europe, and (Halpert and Ropelewski, 1992)

�nd temperature anomalies in South-western Europe and Northern Afri a.

Re ently, new data sets of histori al data have be ome available. This has

opened up the possibility to he k existing onje tures at higher statisti al sig-

ni� an e levels. This study was initiated by a sear h for an ENSO in uen e in

the Netherlands through tropi al storm a tivity. Tropi al storm and hurri ane

a tivity on the Atlanti is suppressed by El Ni~no (Gray, 1984), and many atas-

trophi downpours in De Bilt are remnants of tropi al storms during the At-

lanti hurri ane season. Unfortunately, this does not lead to an observable anti-

orrelation between the NINO3 index and pre ipitation (or high-pre ipitation

events).

In this arti le we present a new analysis of the strongest statisti al on-

ne tion between ENSO and the weather in Europe we did �nd: in reased spring

pre ipitation after an El Ni~no event. In se tion 2 we detail the onne tion for one

station, De Bilt in the Netherlands. The relationship is extended over Europe

in se tion 3, and we sear h for possible me hanisms in se tion 4.

2 Rain in De Bilt

Two years ago we noti ed that there was a orrelation between the strength of

an El Ni~no, quanti�ed by the NINO3.4 index

1

and spring (MAM) pre ipitation

in De Bilt ( entral Netherlands) using data from 1950 to 1995. With a 3-month

lag the linear orrelation oeÆ ient was 0.30, with a nominal signi� an e of 95%.

Given that we had onsidered more than 24 possible relationships, this result

was not very onvin ing. It hinged on one extreme event: in 1983 the spring had

been extraordinarily wet (see Fig. 1). Re ently we used the pre ipitation series

ba k to 1849 and forward to 1998

2

in luding a few more strong El Ni~no events.

Also, Kaplan et al. (1998) give a re onstru tion of the NINO3 index

3

from

1856 to 1991. It uses only sea surfa e temperature (SST) measurements, and

1. the NINO3.4 index isthe average sea surfa e temperature in the region 5

◦S{5

◦N, 120

◦W{

170

◦W, the values where obtained from NOAA/NCEP (Reynolds and Smith, 1994).

2. the series is available upon request from the KNMI.

3. the NINO3 index is the average sea surfa e temperature in the region 5

◦S{5

◦N, 90

◦W{

150

◦W.

2

orrelates quite well with the Jakarta SO index of K�onnen et al. (1998): r = 0.66

in DJF and 0.67 in MAM over 133 years in 1859{1996. From 1950 onwards we

use the NCEP analyses (Reynolds and Smith, 1994), over the overlap period

the orrelation with the Kaplan NINO3 is 0.97. An analysis using the standard

SO index (Allan et al., 1991; K�onnen et al., 1998) gives essentially the same

results.

Figure 1 shows that the relationship between the winter (DJF) NINO3

index and spring (MAM) pre ipitation in De Bilt was on�rmed in the new

analysis. The linear orrelation oeÆ ient is r = 0.35, nominally this has a

han e P < 10−4of being random. The average of the four Dut h stations

with data from 1867 (De Bilt, Groningen, Den Helder and Hoofddorp) gives a

orrelation of 0.40.

To quantize the signi� an e of these relationships we used a Kolmogorov-

Smirnov test to ompute the probability that the pre ipitation distribution

with NINO3 index N3 < N ut

3 is the same as the one with N3 > N ut

3 . The

averages and 2σ bands of these distributions of the De Bilt data are shown in

Figure 2. The di�eren e is signi� ant at the 99% level for N ut

3 > 0.5: an El

Ni~no tends to be followed by a wet spring. On the other hand, the e�e ts of La

Ni~na, in luding the suggestive drought in 1893, are not signi� ant even at the

95% on�den e level at De Bilt. For the four-station average this di�eren e is

also signi� ant at the 95% level for N ut

3 < −0.5.

The signal has no onne tion with the North Atlanti Os illation (NAO).

On the one hand the NAO does not in uen e pre ipitation in the Netherlands

very mu h, on the other hand ENSO and NAO are only orrelated in the sum-

mer.

3 Spring rain in Europe

We investigated the extent of the tele onne tion using the NCDC gridded pre-

ipitation anomalies database (Baker et al., 1995). This ontains global data

from 1851 to 1993 in 5

◦×5

◦bins. In Figure 3 one sees that the spring pre ipita-

tion in reases after an El Ni~no in a zonal belt from England and Fran e to the

Ukraine, with a weaker extension eastwards into Asia. There are two maxmima,

with orrelations oeÆ ients above 0.3 (without the 1998 El Ni~no): southern

England, northern Fran e, the Low Countries and Germany, and another one

in the Ukraine entered on Kiev (r = 0.43). This last point was also noted by

3

r = 0.35

−2 −1 0 +1 +2 +3 +4

DJF NINO3

0

100

200

300

MAM pre .

[mm℄

57

58

5960

61

6263 6465

66

67

68

69

70 71

7273

74

75

76

77

78

79

80

8182

83 84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

00

01

02

03

04

050607

08

09

10

11

1213

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

3132

33

34

35

36

37

38

39

4041

42

43

44

45

46

47

4849

50

51

5253

54

55

56

575859

60

61

62

63

64

65

66

6768

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

Figure 1: A s atter plot of spring (MAM) pre ipitation in De Bilt, Netherlands

versus the NINO3 index of SST in the eastern Pa i� for 1857{1998. Underlined

numbers refer to the 19th entury. The horizontal thin lines give the 10%, 33%,

67% and 90% per entiles.

4

N ut

3

De Bilt

MAM pre .

21.510.50−0.5−1−1.5

300

200

100

0

Figure 2: Mean and 2σ un ertainties of the pre ipitation in De Bilt for the years

with N3 > N ut

3 (dashed urves), and N3 < N ut

3 (solid urves).

Figure 3: The linear orrelation oeÆ ients of the DJF NINO3 index and the

MAM pre ipitation over Europe (1857{1993).

5

r = 0.49

−2 −1 0 +1 +2 +3 +4

DJF NINO3

−60

−40

−20

0

+20

+40

+60

MAM pre .

anomaly

[mm℄

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

8788

89

90

91

9293

9495 96

97

98

99

00

010203

04

0506

07

08

09

1011

1213

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

3940

41

42

43

44

45

46

47

48

49

50

51

52

53

5455

56

57

58

5960

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

8788

89

90

91

92

93

Figure 4: S atter plot of the MAM pre ipitation anomalies in Europe around

50

◦N against the DJF NINO3 index for 1857{1993.

Kiladis and Diaz (1989). In their Fig. 3h a similar but more southerly band

is indi ated over Europe. This band forms a dipole with the drier zone over

Northern Afri a and eastern Spain (r = −0.35), also noted in Kiladis and Diaz

(1989). In ontrast, the orrelation of DJF pre ipitation with the DJF NINO3

index only rea hes values above 0.2 in three grid points: 0.22 in Brussels, 0.23

in Mos ow and −0.29 in Bergen, Norway; none of these rea h 99% signi� an e.

The Iberian signals in the summer and early fall are also weaker.

The relation with ENSO is seen more learly when we onstru t an index

of average MAM rainfall anomalies over the band with positive orrelations

onsisting of the nine 5

◦×5

◦grid boxes entered on 50

◦N from 5

◦W to 35

◦E.

This index has a orrelation oeÆ ient with DJF NINO3 of r = 0.49, and one

an see from �gure 5 that the e�e t now looks signi� ant for all values of N3.

A K-S test on�rms that the distributions are di�erent for the entire range of

6

Europe 50

◦N index

N ut

3

21.510.50−0.5−1−1.5

MAM pre . anom.

80

40

0

−40

Figure 5: Mean and 2σ un ertainties of the pre ipitation anomaly around 50

◦N

from 5

◦W to 35

◦E. Dashed urves: the years with N3 > N ut

3 , solid urves: below

N3 < N ut

3 .

ut-o� values. The orrelation with the NAO is again low, although non-zero

(r = −0.16).

4 Possible mechanisms

Possible me hanisms of this tele onne tion have to explain the time stru ture of

the orrelation. The lag orrelations of the MAM pre ipitation with the NINO3

index is shown in �gure 6. For referen e we also show the DJF orrelations.

Although there is room for an atmospheri me hanism with a time s ale shorter

than a month, the main signal seems to be delayed by 3{6 months. This agrees

with the observation that the orrelations of the MAM NINO3 index with

histori al sea level pressure data (1873{1995, Jones, 1987; Basnett and Parker,

1997) are not very high, though signi� ant. In Fig. 7a one sees that in a band

from the British Isles to the Ukraine sea-level pressure tends to be lower during

El-Ni~no events. The orrelation oeÆ ient r just rea hes −0.20 (P = 97%) over

the North Sea and −0.16 in the Ukraine (P = 92%). It is on average a somewhat

higher in Northern Afri a, r = 0.23 at the Straits of Gibraltar (P = 99%). This

is the orre t stru ture to explain more rain in the dipole of �gure 3. The lag-3

7

Europe 50

◦N index DJF

De Bilt DJF

Europe 50

◦N index MAM

De Bilt MAM

lag [months℄

r

14121086420

0.6

0.5

0.4

0.3

0.2

0.1

0.0

Figure 6: Lag orrelation oeÆ ients of the pre ipitation in spring and winter

in De Bilt and in Europe around 50

◦N with the NINO3 index.

signal (Fig. 7b) has the same features, but is stronger: r = −0.26 over the North

Sea, 0.27 at Gibraltar.

The 3{6 month delay points at the possibility that in addition to this

dire t tele onne tion there is an intermediate variable, probably SST in a third

region, that is in uen ed by ENSO and in turn auses more rain around 50

◦N in

Europe in spring. To investigate this we use the histori al temperature anoma-

lies database of Jones and Parker (Parker et al., 1995; Jones, 1994; Parker

Figure 7: Correlation maps of MAM northern hemisphere SLP (1873-1995) and

MAM NINO3 (a), DJF NINO3 (b), and an index of SE Asia SST ( ).

8

DJF MAM

NINO3 NINO3.4 SE Asia N Pa i� De Bilt 50

◦N

DJF NINO3 1.00 0.80 0.67 0.49 0.35 0.49

MAM NINO3.4 1.00 0.67 0.50 0.22 0.28

SE Asia 1.00 0.47 0.36 0.35

N Pa i� 1.00 0.26 0.30

De Bilt 1.00 0.60

50

◦N 1.00

Table 1: Correlation oeÆ ients between the DJF Kaplan/NCEP NINO3 index;

the MAM NINO3.4, SE Asia and North Pa i� temperature indi es extra ted

from Parker et al. (1995); and MAM De Bilt and European 50

◦N pre ipitation.

et al., 1994), whi h in ludes SST as well as land 2 m temperatures. In Fig. 8

(top panel) we show the orrelation of these temperatures with the European

50

◦N spring pre ipitation index. Lo ally, high pre ipitation is asso iated with

older water in the north-east Atlanti . One also re ognizes the NAO SST sig-

nature in the West Atlanti , in spite of the low orrelation with the atmospheri

NAO index. However, both these patterns are only very weakly asso iated with

ENSO, as one an see from the middle panel in whi h the orrelations of this

MAM temperature �eld and the DJF NINO3 index are shown.

The overlap between these two plots is shown in the bottom panel, in

whi h the produ t of the top two panels is plotted, rlag3NINO3,SST × r

SST,P(50◦N). If

only one area would a t as intermediate variable the lo al value would be equal

to rlag3

NINO3,P(50◦N) = 0.49. Even if more intermediate variables ontribute, areas in

whi h both orrelations are high will stand out in this plot, but a quantitative

interpretation annot be given. One sees that none of the regions rea h values as

high as 0.49, but there are three areas of possible interest: the entral equatorial

Pa i� , south-east Asia and parts of the Indian O ean, and the North Pa i�

dipole.

We de�ned three temperature anomaly indi es orresponding to these re-

gions: for the Central Pa i� we use the NINO3.4 region, for south-east Asia

the box 60

◦E { 120

◦E, 10

◦S { 20

◦N and for the North Pa i� dipole the region

160

◦E { 120

◦W, 30

◦N { 60

◦N with a top-left/bottom-right dipole stru ture.

The orrelation oeÆ ients between all indi es we a umulated are given in Ta-

ble 1. In �gure 9 we plot the lag- orrelation oeÆ ients of these indi es with the

NINO3 index, multiplied by their orrelation with the European 50

◦N spring

9

rSST,P(50◦N)

rlag3NINO3,SST

rlag3NINO3,SST × r

SST,P(50◦N)

Figure 8: The orrelation of spring pre ipitation around 50

◦N in Europe with

the Jones and Parker temperature dataset (top), the lag-3 orrelation of this

temperature and the NINO3 index (middle) and the produ t of the these two

orrelations (bottom). Note the di�erent s ales.

10

lag [months℄

r

0.30 ×N Pa i�

0.35 × SE Asia

0.28 × NINO3.4

14121086420

0.3

0.2

0.1

0.0

Figure 9: Lag orrelation oeÆ ients of the MAM temperature anomalies in the

three possible sour e regions with the with the NINO3 index, weighted with

the orrelation oeÆ ients with the Europe 50

◦N pre ipitation index in spring.

pre ipitation index. One sees that the shape of the lag stru ture of the south-

east Asian region orresponds losely to the observed signal. Also the signal

in pressure (Fig. 7 ) is very similar to the lagged NINO3 signal over Europe

(Fig. 7b), but the North-Afri an opposite-sign anomaly has disappeared. How-

ever, the orrelation of the south-east Asia index with rainfall in Europe is

only 0.35, whi h is lower than the lagged orrelation with NINO3 instead of

the higher value expe ted for an intermediate variable. This may partly be due

to other me hanisms, su h as the dire t link from the entral Pa i� , but also

the nature of the measurements plays a role. Temperature di�eren es in the

south-east Asia area are small: the MAM varian e is only (0.28 K)2, mu h less

than the (0.88K)2of the DJF NINO3 index. This in reases the e�e t of noise in

the form of measurement errors, irrelevant land points and small-s ale weather.

On the basis of the time delays of the signal we on lude that south-east Asia is

most likely the main sour e of the in uen e of ENSO on the weather of Europe,

with a smaller dire t link from the entral Pa i� .

The importan e of the south-east Asia sea surfa e temperature for the

northern hemisphere ir ulation is supported by data analyses and modelling

studies, see e.g. the review by Trenberth et al. (1998). In spring the most a tive

11

tele onne tion is the North Pa i� pattern. This is the top half of Figs 7a,b.

The lower halves show an extension a ross the North Pole into Europe. This

extension substantiates arguments using simpli�ed Rossby wave propagation.

Unfortunately virtually all modelling studies have been ondu ted for northern

winter onditions, when the observations ontain a mu h weaker tele onne tion.

Still, indi ations of the pole- rossing response an be seen in DJF AGCM results

(see, e.g., Ferranti et al., 1994). One ould spe ulate that the spring transition to

the Asian and Chinese monsoon systems makes the ir ulation more sus eptible

to perturbations. Further modelling work is learly needed in order to elu idate

the me hanism behind the tele onne tion to Europe.

5 Conclusions

Using more than a entury of data a lear in uen e of ENSO on the weather

in Europe has been established: spring pre ipitation in a belt around 50

◦N

from Southern England to the Ukraine tends to in rease after an El Ni~no and

de rease after a La Ni~na. The strength of the orrelation is r = 0.49 for an index

of pre ipitation in this belt, r = 0.40 for the average of four Dut h stations,

and r = 0.35 for the single station De Bilt. Other European tele onne tions

are weaker than this. The 3-6 month lag and orrelation maps suggest that

south-east Asian surfa e temperatures may a t as an intermediate variables for

most of the signal.

A knowledgements We would like to thank Theo Opsteegh and Robert Mureau

for many useful dis ussions.

References

Allan, R.J., Ni holls, N., Jones, P.D., and Butterworth, I.J. 1991. `A further

extension of the Tahiti-Darwin SOI, early SOI results and Darwin pressure',

J. Climate, 4, 743{749. The series an be found at http://www. ru.uea.a .uk/

ru/data/soi.htm.

Baker, C. B., Eis heid, J. K., Karl, T. R., and Diaz, H. F. 1995. `The quality

ontrol of long-term limatologi al data using obje tive data analysis', in

Preprints of AMS Ninth Conferen e on Applied ClimatologyDallas, TX. The

12

data are available from http://www.n d .noaa.gov/onlinedata/ limatedata/

grid.pr p.seasanom.html.

Basnett, T. A. and Parker, D. E. 1997. `Development of the global mean sea level

pressure data set GMSLP2', Climati Resear h Te hni al Note 79 Hadley

Centre Meteorologi al OÆ e, Bra knell, U.K. 16pp plus Appendi es. Data

are available from http://www. ru.uea.a .uk/ ru/data/pressure.htm.

Berlage, H. P. 1966. The Southern Os illation and World Weather Number 88

in Mededelingen en verhandelingen. KNMI 152 pp.

Ferranti, L., Molteni, F., and Palmer, N. 1994. `Impa t of lo alized tropi al and

extratropi al SST anomalies in ensembles of seasonal GCM integrations', Q.

J. Meteorol. So ., 120, 1613{1645.

Fraedri h, K. 1994. `An ENSO impa ty on Europe?', Tellus, 46A, 541{552.

Gray, W. M. 1984. `Atlanti seasonal hurri ane frequen y: Part I: El Ni~no and 30

mb quasi{biennial os illation in uen es', Mon. Wea. Rev., 112, 1649{1668.

Halpert, M. S. and Ropelewski, C. F. 1992. `Surfa e temperature patterns as-

so iated with the Southern Os illation', J. Climate, 5, 577{593.

Jones, P. D. 1987. `The early twentieth entury ar ti high | fa t or � tion?',

Climate Dynami s, 1, 63{75.

Jones, P. D. 1994. `Hemispheri surfa e air temperature variations: a reanalysis

and an update to 1993', J. Climate, 7, 1794{1802.

Kaplan, A., Cane, M., Kushnir, Y., Clement, A. C., Blumenthal, and Ra-

jagopalan, B. 1998. `Analyses of global sea surfa e temperature 1856{1991',

J. Geophys. Res., 103, 18567{18589. Data are available from http://

ingrid.ldgo. olumbia.edu.

Kiladis, G. N. and Diaz, H. F. 1989. `Global limati anomalies asso iated with

extremes in the southern os illation', J. Climate, 2, 1069{1090.

K�onnen, G. P., Jones, P. D., Kaltofen, M. H., and Allan, R. J. 1998. `Pre-

1866 extensions of the Southern Os illations index using early Indonesian

and Tahitian meteorologi al readings', J. Climate, 11, 2325{2339.

Parker, D. E., Folland, C. K., and Ja kson, M. 1995. `Marine surfa e temper-

ature: observed variations and data requirements', Climati Change, 31,

559{600.

Parker, D. E., Jones, P. D., Bevan, A., and Folland, C. K. 1994. `Interde adal

hanges of surfa e temperature sin e the late 19th entury', J. Geophys.

Res., 99, 14373{14399. Data are available from http://www. ru.uea.a .uk/

ru/data/temperat.htm.

Reynolds, R. W. and Smith, T. M. 1994. `Improved global sea surfa e analyses

13

using optimum interpolation', J. Clim., 7, 929{948. Nino indi es are available

from the Climate Predi tion Center at http://ni .fb4.noaa.gov/data/ ddb/

altindex.html.

Trenberth, K. E., Branstator, G. W., Karoly, D., Kumar, A., Lau, N.-C., and

Ropelewski, C. 1998. `Progress during TOGA in understanding and modeling

global tele onne tions asso iated with tropi al sea surfa e temperatures', J.

Geophys. Res., 103, 14291{14324.

van Loon, H. and Madden, R. A. 1981. `The Southern Os illation, Part I. Global

asso iations with pressure and temperature in northern winter', Mon. Wea.

Rev., 109, 1150{1162.

Wilby, R. 1993. `Eviden e of ENSO in the synopti limate of the British Isles

sin e 1880', Weather, 48, 234{239.

14