Indonesian Journal of Physics

Kontribusi Fisika Indonesia

Vol. 14 No.3, Juli 2003

The Spectrum Analysis of Rainfall in Indonesia

Sinta Berliana Sipayung, Hariadi T.E, Nurzaman A dan Eddy Hermawan

Pusat Pemanfaatan Sains Atmosfer dan Iklim LAPAN
Jl. Dr. Junjunan 133, Bandung 40173
[email protected]

Abstract
As a specific region at equatorial, namely maritime continent, the characteristics of meteorological element,
especially rainfall in Indonesia is influenced by monsoon climate, called typical rainy and dry season. We have
analyzed monthly mean of rainfall measured at twenty two (22) stations over Aceh, Medan, Padang, Jambi up to
Biak for the period of 20 years (1970 to 1989). By using bandpass filter and spectrum analysis we could obtain the
characteristics and predominance of Indonesian rainfall oscillation. We found that the characteristics of these
rainfall could be identified in form of Semi-Annual (SAO), Quasi-Biennial Oscillation (QBO), and EL-
Nino/Southern Oscillation (ENSO). We also could identify the predominance peak of these oscillations.

Keywords : Spektrum Analysis, Annual Oscillation

Abstrak
Sebagai satu-satunya kawasan khusus di daerah katulistiwa yang dikenal sebagai benua maritim, karakteristik dari
unsur-unsur meteorologi khususnya curah hujan di atas wilayah Indonesia sangat dipengaruhi oleh kondisi iklim
monsun yakni musim basah dan musim kering. Kami telah menganalisa curah hujan rata-rata bulanan hasil
pengukuran di dua puluh dua (22) stasiun penakar curah hujan yang tersebar di seluruh wilayah Indonesia mulai
dari Aceh, Medan, Padang, Jambi hingga Biak selama 20 tahun pengamatan (1970 hingga 1989). Dengan
menggunakan band-pass filter dan analisis spektrum, kami dapat mengidentifikasi karakteristik curah hujan di
Indonesia kedalam bentuk Semi-Annual Oscillation (SAO), Annual Oscillation (AO), Quasi-Biennual Oscillation
(QBO) dan El-Nino/Southern Oscillation (ENSO). Kami juga dapat mengidentifikasi puncak yang dominan dari
masing-masing osilasi tadi.

Kata kunci : Analisa Spektrum, Osilasi tahunan.

northern Pacific Ocean, together with cold surges

1. Introduction
from China to the southern hemisphere. This cold
Globally there are three equatorial heat sources surges bring much of water vapor. The Coriolis force
that are interpreted as the driving force for the general swiches the wind direction in the southern
circulation. One is Indonesia region which is called hemisphere. The wind blows through Sumatera and
the “maritime continent”1), and other two are located converges to Java island.
over Africa and South America. Indonesia region is The summer monsoon appears when the
located between two large (great) continents, Asia and depression over Asia, especially south China and
Australia, and between two oceans, Pacific ocean and Tibetan Plateau, is developed. The wind blows from
Indian ocean. Consequently, the surface southern hemisphere to the northern hemisphere
meteorological elements in these regios are during summer monsoon and brings much water vapor
characterized by “monsoon climate”. to the southeast Asia. During this season, Indonesia
The monsoon variability should be considered region is influenced by the wind from Australian dry
as part of the oscillation in the coupled ocean/land/ airmass. After passing the equator, the Coriolis force
atmosphere system, which involves the largest changes the wind direction so that the winds becomes
continent and the largest ocean. This monsoon climate westerly in the northern hemisphere.
has typical rainy and dry season. The rainy season It has been well documented that the
corresponds to the northeast monsoon period in the interannual variability of monthly rainfall in Indonesia
northern winter (winter monsoon), and the dry season are strongly related to the ENSO4-8).
comes to the southwest monsoon period in the
2. Data
northern summer (summer monsoon) as explained by
Yasunary2). Table 1 shows data observed at 22 provinces in
Figures 1. a and b, a primary synoptic-scale Indonesia since 1970 to 1989. These data were
circulation and associated precipitation features on the obtained from Data Iklim Indonesia published by the
Asian winter and summer monsoon are studied by Institute of Meteorology and Geophysics, Department
Johnson3). We can see the cold surges moves from the of Communication of Indonesia.

97
98 KFI Vol. 14 No. 3, 2003

The locations of the cities are shown in Figure Annual Oscillation (AO), Quasi Biennial Oscillation
2 while the time series of rainfall at Padang and (QBO) and El Niño and Southern Oscillation (ENSO).
Surabaya are shown in Figures 3.a dan 3b as a sample In our filter, the band passed output can be obtained
characteristics of rainfall in Indonesia. From these by Maruyama
figures we can see that the characteristics of the Yk = a(xk – xk-2) – b1yk-1 – b2yk-2
rainfall at Padang is different from Surabaya. This is a
sample only that the characteristic of rainfall in (k = 1,2,……, N) (3.1.1)
Padang relatively is different with in Surabaya since where (x1, x2, ……,xN ) show N input data series, and
this region is located close to equator. While, y1, y2 ,……, yN ) show the corresponding N output data
Surabaya is located a little far way from equatorial. series.
The mean value of rainfall at Padang is relatively In order to specify the coefficients a, b1 and b2,
larger than Surabaya. A similarly, the oscillation we use three frequencies ω 0, ω 1 and ω 2 (measured
amplitude of rainfall at Padang is relatively larger than by 2 Π / period). ω 0 is the central frequency with
at Surabaya. It shows the periodic variation of rainfall full-value response, and ω 1 and ω 2 are the
at Padang seems more complex than at Surabaya also frequencies with half-value response on both sides of
(see Figs.3. a and b). ω 0. There is a relation between these three
3. Methods of Data Analysis. parameters, as follow,
3.1. Filtering Method ω 02 = ω 1 ω 2. (3.1.2)

We used recursion technique for band-pass

filter to indicate Semi Annual Oscillation (SAO),

Table 1. Data observed at 22 province in Indonesia.

No. Station Name Region Location

Lat. Lon.
1 Aceh Sumatera Utara 5.52 95.19
2 Medan Sumatera Utara 3.37 98.48
3 Padang Sumatera Barat -0.53 100.21
4 Tanjung Pinang Sumatera Barat 0.55 104.32
5 Pangkal Pinang Sumatera Barat -2.10 106.08
6 Jambi Sumatera Selatan -1.38 103.39
7 Lampung Sumatera Selatan -4.27 105.11
8 Jakarta Jawa Barat -6.10 106.10
9 Bandung Jawa Barat -7.01 107.17
10 Jogjakarta Jawa Tengah -7.78 110.43
11 Borobudur Jawa Tengah -7.07 110.01
12 Surabaya Jawa Timur -7.03 113.58
13 Pontianak Kalimantan Barat -0.01 109.23
14 Balikpapan Kalimantan Timur -1.16 116.54
15 Gorontolo Sulawesi Utara 0.38 122.50
16 Luwuk Sulawesi Tengah -0.45 122.47
17 Ujung Pandang Sulawesi Selatan -5.04 119.33
18 Saumlaki Maluku -7.59 131.18
19 Ampenan Nusa Tenggara Barat -8.32 116.04
20 Maumere Nusa Tenggara Timur -8.28 122.21
21 Denpasar Bali -8.46 115.10
22 Biak Irian Jaya -1.11 136.07

2(Ω 02 − 4)
In this study, we used 6,12,27, and 52 months period b1 = (3.1.4)
as central frequencies of SAO, AO, QBO and ENSO, 4 + 2∆Ω + Ω 02
respectively. And the sampling interval ∆ T is taken
to be 1 month. Here, a,b1 and b2 can be specified as : 4 − 2∆Ω + Ω 02
b2 = (3.1.5)
2∆Ω 4 + 2∆Ω + Ω 02
a = (3.1.3)
4 + 2∆Ω + Ω 02 where ∆Ω and Ω 0 are obtained as
⎡ sin ω1∆T sin ω2 ∆T ⎤
∆Ω = 2 ⎢ − ⎥ (3.1.6)
⎣1 + cos ω1∆T 1 + cos ω2 ∆T ⎦
KFI Vol. 14 No. 3, 2003 99

⎡ 4 sin ω1 ∆T sin ω 2 ∆T ⎤ The power spectrum of data YI similarly is given by

∆Ω = 2 ⎢ ⎥ (3.1.7)
⎣ (1 + cos ω1 ∆T )(1 + cos ω 2 ∆T ) ⎦
M
Pk (Y) = 2∆t ∑ (Y , Y ) klπ
L Dl ∂l cos M
l =0
cl

The actual calculation consists of two steps in

the same manner as Shanks9) and Murakami10). At (k=0,1,2,…….M). (3.2.9)
first, setting y1 and y2 to be zero, the temporary output 4. Results
data are calculated by applying Eq. (3.1.1) to the
input data from which the mean value and linear Filtering and power spectrum methods are
trend have been removed. Then, this temporary applied to monthly mean data obtained at the stations,
output data are reversed with time in Eq. (3.1.1); k = 22 province for the period of 20 years (1970 to 1989).
N, N-1,……, 1. By using band-pass filter and spectrum analysis
methods we found four oscillation patterns with
3.2 Spectrum Method predominant peaks that appear in each province :
The spectral analysis in this study is referred 1. Semi-Annual Oscillation (SAO)
to Maruyama11). The covariance or the lag correlation 2. Annual Oscillation (AO)
between two time data Xi and YI are given by 3. Quasi-Biennial Oscillation (QBO)
N 4. El-Nino/Southern Oscillation (ENSO)
∑
1
Ll (X,Y ) = XI YI-l
N − l − m i = l +1 4.1 Filtered Data
(l+0,1,2,……….,M) (3.2.1) SAO, AO, QBO and ENSO oscillations are
approximately obtained by using band-pass filter for
where l is the lag number, N is the total number of
5-7 months, 10-14 months, 22-32 months, and 40-66
time series data, m is the number of undefined
months, respectively, and are shown in Figs 4. Band-
missing data, and M is the maximum lag number.
pass filtering of rainfall data shows that amplitude of
Even and odd parts of the covariance between
rainfall variation at Padang relatively is larger in
X and Y are given by
SAO than at Surabaya (Figs. 4. a and b), but smaller
1
Lcl (X,Y) = [Ll (Y,X) + Ll (X,Y)] (3.2.2 ) in AO (Figs. 4.c and d). For QBO, the amplitude of
2 rainfall variation at Padang is a little larger than at
Surabaya (Figs. 4. e and f). So, we can mention that
1
Lsl (X,Y) = [Ll (Y,X) - Ll (X,Y)] (3.2.3) rainfall at Padang is dominated by SAO and that
2 rainfall at Surabaya is dominated by AO. And for
The power spectra are given by correlations ENSO, it is pretty difficult to find the amplitude of
between even/odd parts of the covariance and rainfall variation at the two cities (see Figs. 4. g and
cosine/sine, respectively11). By use of time series data h). In this we do not show the band-pass filter for all
given at interval of ∆t, spectral values are determined of stations, for Padang and Surabaya only.
within the range of 0 to ½ ∆t. The maximum lag 4.2 Spectra
number M gives M + 1 spectral estimates of intervals
of ½ M ∆t. Spectral analysis based on the Blackman-
To make the covariance smoothly tend to zero Tukey method11) was applied to the monthly mean
at the edge of lagging (l = 0, M) a lag window is data of rainfall for the period from 1979 1989
applied by observed at 22 province. As is expected from time
variation of filtered data described in sub section 4
lΠ
Dl = cos2 (3.2.4) existence of periodic oscillations of SAO, AO, QBO,
2M and ENSO can be confirmed by appearance of power
The power with frequency (k/2M) ∆t therefore is peaks as Figs of 5 to 8.
given by Figure 5. a. shows that the most predominant
M
klΠ peak of rainfall at Aceh province is AO, and the
p ( X ) = ∑ Lcl ( X , X ) Dl δ l cos (3.2.5)
second is SAO, and third is QBO. At Medan, the
k
l =0 M
most predominant peak of rainfall is AO, and the
where δ l is the width factor of lagging is given by second is SAO, and third is ENSO and than QBO
(Fig. 5. b). But different with Padang (Fig. 5.c) the
δ l = ½ (l=0,M) and δ l = 1 (l=0,M) (3.2.6) most predominant peak of rainfall is SAO, and the
Thus, we obtain spectral values of power second is AO, and third is QBO. Note here that SAO
density, that is, power per unit frequency: and AO are almost the same in power. The same
fenomena appear at Tanjung Pinang (Fig. 5. d)
Pk (X) =pk (x) 2 M ∆T, (3.2.7) predominant peak is SAO, and the second is AO, and
That is, third is ENSO.
M klπ Fig. 5. e shows that the most predominant
Pk (X) = 2∆t ∑ ( X , X )
Lcl Dl ∂l cos
l =0
M
peak of rainfall Pangkal Pinang is AO and the other
fenomenon is very weak and is difficult to detect any
significant peak, at Jambi is AO, and the second is
(k=0,1,2,……..M). (3.2.8)
100 KFI Vol. 14 No. 3, 2003

SAO, and the third is ENSO (Fig 5. f). Also at 2. Yasunari, T., Temporal and spatial variations of
Lampung Fig. 5. g, Jakarta Fig. 6. h, Bandung Fig. 6. rainfall in Java, Indonesia, Southeast Asian
i, Jogjakarta Fig. 5. j, Borobudur Fig. 5. k and Studies, 19, 170-186, (1981).
Surabaya Fig. 5. l. 3. Jhonson, R. H., Heat and moisture sources and
Rainfall at Pontianak is dominated by SAO sinks of Asian monsoon precipitating system, J.
and AO, and the third and four predominant peaks Meteor. Soc. Japan, 70, 353-371, (1991).
appear are QBO and ENSO (see Fig. 5. m). At 4. Berlage, H. P., Fluctuations in the general
Balikpapan, the most predominant peak is SAO, and atmospheric circulation of more than one year,
the second is ENSO but very weak as shown in Fig. their nature and prognostic value,
5. n. Verhandelingen, 69, 152pp, (1957).
Fig. 5. o shows that the most predominant 5. Berlage, H. P., The southern oscillation and
peak at Gorontalo is SAO, and the second is ENSO, word weather, Verhandelingen, 88, 152pp,
and different at Sulawesi Tengah Fig. 5. p are (1966).
dominated AO and ENSO. 6. Hackert, E. C & S. Hastenrath., Mechanisms of
At Ujung Pandang Fig. 5. q, Maluku Fig. 5. r, Java rainfall anomalies, Mon. Wea. Rev., 114,
Ampenan Fig. 5. s and NTT Fig. 5. t the most 745-757, (1986).
predominant peak is AO, the others fenomena are 7. Karyoto, Haryanto, H., Sasmito, A., Sumiratno,
very difficult to detect. But at Bali is dominated by Santoso, E., Nuryadi, & Antoyo., Climate
AO and QBO Fig. 5. u, and at Biak is very strong variation in Indonesia and their relation to
each fenomena (Fig. 5. v). ENSO phenomena, 4th ICEAR Symp. On
As already investigated by Sipayung. S. B and Equatorial Atmosphere Observations over
Tanaka. H12), they found that Jakarta and Surabaya Indonesia, 10-11 Nov. 1992. Jakarta, Indonesia,
have the same rainfall pattern while Padang has (1992).
different pattern. 8. Yasunari, T. & R. Suppiah., Some problem on
the Interannual variability of Indonesian
5. Conclusions
monsoon rainfall, Tropical Rainfall
This study is mainly concerned an application Measurement, Deepak Publishing, 113-121,
of spectrum analysis on rainfall data over 22 stations 1988.
in Indonesia. We found that : Although the most 9. Shanks, J. L., Recursion filters for digital
predominant peak of rainfall over Indonesia is processing, Geophysica, 32, 33-51, (1967).
dominated by AO, but some parts of these region is 10. Murakami, M., Large-scale aspects of deep
dominated by SAO such as Padang, Tanjung Pinang convective activity over the GATE area. Mon,
and Kalimantan. We also found ENSO phenomena, Wea. Rev., 107, 994-1013, (1979).
but this is not a predominant peak. They are Tanjung 11. Maruyama, T., Time sequence of power spektra
Pinang, Jambi, Pontianak, Balikpapan, Gorontalo, of disturbances in the equatorial lower
Sulawesi Tengah dan Biak. We need more data stratosphere in relation to the QBO, J. Meteor.
analysis to get more informations of the Soc. Japan, 46, 327-341, (1968).
characteristics of rainfall over Indonesia. 12. Sipayung S. B & Tanaka. H., The spektrum
analysis of meteorological element in Indonesia,
Master thesis in Universitas Nagoya, Japan,
References 1995.
1. Ramage, C. S., Role of a tropical “Maritime
Continent” in the atmosferic circulation, Mon.
Wea. Rev., 96, 365-370, (1968).
KFI Vol. 14 No. 3, 2003 101

(a)

(b)

Figure 1. Primary circulation features that affect cloudiness precipitation in the region of the winter
monsoon (a). Primary circulation features that affect cloudiness and precipitation in the region of the
summer monsoon (b)3)
102 KFI Vol. 14 No. 3, 2003

Figure 2. Indonesian map which showing the location of radiosonde station over Indonesia.

Rainfall in Padang Rainfall in Surabaya

1200
600
1000
500
Rainfall(mm)
Rainfall(mm)

800 400
600 300
400 200

200 100
0
0
1 25 49 73 97 121 145 169 193 217 1 24 47 70 93 116 139 162 185 208 231
Monthly Monthly

(a) (b)

Figure 3. The time series of rainfall in Padang and Surabaya. The Characteristics of rainfall in Padang
is different from Surabaya, mean value at Padang is larger than Surabaya.
KFI Vol. 14 No. 3, 2003 103

Figure 4. Band-pass filtered data of rainfall at Padang and Surabaya. Arrows in g and h denote El-Nino events.
104 KFI Vol. 14 No. 3, 2003

Rainfall in Aceh Rainfall in Medan

800000 300000
700000
Power(mm /month)

250000

Power(mm /month)
600000
500000 200000

400000 150000
300000
100000
200000
100000 50000

0
0
0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 140

Period(month) Period(month)

a) b)

Rainfall in Padang Rainfall in Tanjung Pinang

400000 200000
350000 180000
Power(mm /month)

Power(mm /month)
160000
300000
140000
250000 120000
200000 100000
80000
150000
60000
100000 40000
50000 20000
0
0
0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 140

Period(month) Period(month)

c) d)

Rainfall in Pangkal Pinang Ranfall in Jambi

450000 200000
400000
Power(mm /month)

Power(mm /month)

350000 150000
300000
250000
100000
200000
150000
50000
100000
50000
0
0
0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 140

Period(month) Period(month)

e) f)

Figure 5. The spectral power density for rainfall at Aceh, Medan, Padang, Tanjung Pinang, Pangkal Pinang dan Jambi. The
most predominant peak at Aceh, Medan, Pangkal Pinang and Jambi are Annual Oscillation but at Padang and Tanjung
Pinang are Semi-annual Oscillation.
KFI Vol. 14 No. 3, 2003 105

Rainfall in Lampung Rainfall in Jakarta

1000000 900000
900000 800000
Power(mm /month)

800000

Power(mm /month)
700000
700000
600000
600000
500000
500000
400000 400000
300000 300000
200000 200000
100000 100000
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140

Period(month) Period(month)

(g) (h)

Rainfall in Bandung Rainfall in Yokja

1600000 2000000
1400000 1800000
Power(mm /month)

1600000
1200000
1400000
1000000 1200000
800000 1000000
600000 800000
600000
400000
400000
200000 200000
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140

Period(month) Period(mont h)

(i) (j)

Rainfall in Borobudur Rainfall in Surabaya

1200000
1400000
Spektral(mm /month)

Power(mm /month)

1200000 1000000
1000000 800000
800000
600000
600000
400000 400000

200000 200000
0
0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
Period(month)
Period(month)

(k) (l)

As the same Fig. 5., but for Lampung, Jakarta, Bandung, Yogyakarta, Borobudur and Surabaya. The most predominant
peak at Lampung, Jakarta, Bandung, Jogjakarta Borobudur and Surabaya are Annual Oscillation.
106 KFI Vol. 14 No. 3, 2003

Rainfall in Pontianak Rainfall in Balikpapan

300000 160000
140000
Power(mm /month)

250000

Power(mm /month)
120000
200000
100000
150000 80000
100000 60000
40000
50000
20000
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140

Period(month) Peiod(month)

(m) (n)

Rainfall in Gorontalo Rainfall in Luwuk

60000 120000
Power(mm /month)

Power(mm /month)

50000 100000

40000 80000

30000 60000

20000 40000

10000 20000

0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
Period(month) Period(month)

(o) (p)

Rainfall in Ujung Pandang Rainfall in Saumlaki

10000000 1000000
900000
8000000
Power(mm /month)

800000
700000
6000000
600000
500000
4000000
400000
2000000 300000
200000
0 100000
0 20 40 60 80 1 00 1 20 1 40 0
0 20 40 60 80 100 120 140

Period(mont h) Period(month)

(q) (r)

As the same as Fig. 5, but for Pontianak, Balikpapan, Gorontolo, Sulawesi Tengah, Ujung Pandang dan Maluku. The most
predominant at Pontianak, Balik Papan and at Gorontalo are Semi- Annual Oscillation but for Luwuk, Ujung Pandang and
Saum Laki are Annual Oscillation.
KFI Vol. 14 No. 3, 2003 107

Rainfall in Ampenan Rainfall in Naumere

800000 900000
700000 800000
Power(mm /month)

Power(mm /month)
600000 700000

500000 600000

400000 500000
400000
300000
300000
200000
200000
100000
100000
0
0
0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 140

Period(mm /month) Period(month)

(s) (t)

Rainfall in Denpasar Rainfall in Biak

2000000
1800000 60000
Power(mm /month)

1600000 50000
Power(mm /month)

1400000
40000
1200000
30000
1000000
800000 20000
600000 10000
400000
0
200000
0 20 40 60 80 100 120 140
0
0 20 40 60 80 100 120 140
Period(month)
Period(month)

(u) (v)

As the same as Fig. 5, but for Ampenan, Maumere, Denpasar and Biak. At Ampenan, Maumere and Denpasar the most
predominanat peak are Annual Oscillation, in Biak there are many peak oscillation where ENSO is the most predominant
peak oscillation following by Annual Oscillation. We suspect because Biak is located close to Pacific Ocean.