7732 A.R. As-syakur et al.
Figure 9. Spatial distribution of peak amplitude phases months derived from TRMM 3B43:
a maximum and b minimum phases.
has two peaks of heavy rainfall Aldrian and Susanto 2003. These phenomena were not supported in this study.
In Papua Island, the peak phase of maximum rainfall occurred in MAM and the min- imum during JJA see Figure 9a. In addition to the ITCZ, these conditions are also
influenced by the effects of the northwest monsoon and a large-scale zonal east–west circulation over the equatorial Pacific Hall 1984. In Maluku Islands, the peak rainfall
phase maximum differs from that in other areas during JJA, in line with the western tropical Pacific region. The climate in this region is affected by the conditions of the western tropical
Pacific region Aldrian and Susanto 2003; Kubota et al. 2011. The reason for this influ- ence is unclear, although Aldrian and Susanto 2003 explain that during JJA the Indonesian
throughflow ITF brings warm water from a warm pool area in the western tropical Pacific region, resulting in heavy rainfall during this season. During the dry season in these areas,
cooler water moving from the warm pool to the Maluku Sea inhibits the formation of a convective zone and results in a minimum rainfall peak phase see Figure 9b.
Figure 10 shows an increasing trend of spatial distribution for monthly rainfall time- series data from January 1998 to December 2010 collected by TRMM 3B43 over Indonesia.
In general, rainfall trend was more positive in the north than the south of Indonesia. In addi- tion, rainfall tended to increase over land areas, especially the four major islands, except
Sulawesi Island. On the other hand, downward trends in rainfall occurred along the western and southern coast of Sumatra, eastern Jawa, southern Sulawesi, Maluku Islands, west-
ern Papua, and Bali Island. The extreme climate of ENSO and IOD was associated with interannual months of Indonesian rainfall variability. For 13 years, several extreme rainfall
events were influenced by ENSO and IOD, causing variations in rainfall trends.
4.2. Comparison with rain gauges
The monthly TRMM 3B43 satellite data products were compared with gauge observa- tions from the five stations over Indonesia see Figure 1, and we sought to determine the
3B43 estimated values and the magnitude of rainfall on the ground. The average results Downloaded by [103.29.196.19] at 07:45 03 September 2013
International Journal of Remote Sensing 7733
90 °
100 °
110 °
120 °
130 °
140 °
150 °
90 °
100 °
110 °
120 °
130 °
140 °
150 °
−0.045 to −0.035 Trend
mm month
−1
0.035 −0.046
0.025 −0.035
0.015 −0.025
0.005 −0.015
−0.005−0.005 −0.015 to −0.005
−0.025 to −0.015 −0.035 to −0.025
10 °
10 °
0° 10
°
10 °
°
Figure 10. Spatial-distribution trend of time-series monthly rainfall from TRMM 3B43 from
January 1998 to December 2010.
1000 2000
3000 4000
Medan Pontianak Denpasar Ternate
Jayapura Rain gauges location
Rainfall mm year
− 1
Rain gauge 3B43
Figure 11. Histograms of annual rainfall between 3B43, with rain gauge estimates from five
locations, over the period from 1998 to 2010.
from point-by-point analysis across all stations at five rain gauges show that annual rain- fall from 3B43 was lower than that from the gauge data: 2414.38 and 2470.81 mm year
−1
, respectively. However, in regard to individual rain gauges, the amount of rainfall varied.
Figure 11 shows histograms of annual rainfall from 3B43 and rain gauge estimates in five locations during 1998–2010. The histograms of annual rainfall indicate that there are two
rain gauges Medan and Pontianak in the western part of Indonesia with values lower than satellite estimates overestimated, whereas three rain gauges Denpasar, Ternate, and
Jayapura in the eastern part of Indonesia have rainfall values higher than satellite esti- mates underestimated. Overestimated values were also found in Malaysia western part
of Indonesia by Semire et al. 2012, and underestimated in Bali eastern part of Indonesia by As-syakur et al. 2011.
Figure 12 shows the intra-annual variation of the long-term mean monthly rainfall mea- sured by 3B43 and the rain gauge data from five stations. This figure indicates that the
similarity of monthly rainfall patterns from five locations confirmed the close relationship between 3B43 and rain gauges. The satellite data and ground reference data yielded high
to very high correlations for these products for each rain gauge. The correlations between 3B43 and rain gauges in Medan, Pontianak, Denpasar, Ternate, and Jayapura are 0.98, 0.90,
0.98, 0.95, and 0.85, respectively.
The comparison histograms of peak maximum, peak minimum, and amplitude variabil- ity derived from TRMM 3B43 and rain gauges from five locations are shown in Figure 13.
Downloaded by [103.29.196.19] at 07:45 03 September 2013
7734 A.R. As-syakur et al.
100
a
d e
c b
200 300
400 500
Ja n
Fe b
Ma r
Ap r
Ma y
Ju n
Ju l
Au g
Se p
Oc t
No v
De c
Month R
ain fa
ll mm mo n
th
− 1
Ja n
Fe b
Ma r
Ap r
Ma y
Ju n
Ju l
Au g
Se p
Oc t
No v
De c
Month Ja
n Fe
b Ma
r Ap
r Ma
y Ju
n Ju
l Au
g Se
p Oc
t No
v De
c Month
100 200
300 400
500
Ja n
Fe b
Ma r
Ap r
Ma y
Ju n
Ju l
Au g
Se p
Oc t
No v
De c
Month Ra
in fa
ll m
m m
o n
th
− 1
Ja n
Fe b
Ma r
Ap r
Ma y
Ju n
Ju l
Au g
Se p
Oc t
No v
De c
Month Rain gauge
3B43
Figure 12. Monthly average rainfall pattern measured by 3B43 and rain gauges. a Medan,
b Pontianak, c Denpasar, d Ternate, and e Jayapura, over the period from 1998 to 2010.
100 200
300 400
500
M ax
imu m
Mi n
im u
m V
ar ia
b ility
M ax
imu m
M in
imu m
V ar
ia b
ility M
ax imu
m Mi
n im
u m
V ar
ia b
ility M
ax imu
m M
in imu
m V
ar ia
b ility
M ax
imu m
M in
imu m
V ar
ia b
ility Medan
Pontianak Denpasar
Ternate Jayapura
Rain gauges location R
ai n
fa ll
mm mo n
th
− 1
Rain gauge 3B43
Figure 13. Comparison histograms of maximum, minimum, and amplitude variability averaged
rainfall measured by 3B43 and rain gauges from January 1998 to December 2010.
The histograms between peak maximumminimum rainfall phase amplitude and annual rainfall show a similar pattern; two rain gauges have values lower than satellite estimates
and three rain gauges have rainfall values higher than satellite estimates. Peak maximum and amplitude variability display great differences between satellite and rain gauge data
compared with the peak minimum rainfall phase. Overall, a comparison between monthly 3B43 products and gauge observations show that differences still exist.
We suggest that the reasons for these differences between 3B43 and rain gauge data are due to temporal and spatial sampling uncertainties. First, since TRMM is a major
component of the 3B43 product, one would expect it not to record some rainfall owing to unfavourable timing of its overflight of Indonesia. Fleming et al. 2011 found this to
be the case in a study in Australia. Because of the non-Sun-synchronous satellite orbit, the TRMM records locations approximately once every 3.6 days in the tropical region
Downloaded by [103.29.196.19] at 07:45 03 September 2013
International Journal of Remote Sensing 7735
As-syakur 2011. Furthermore, limitations of the TRMM data suffer from both a narrow swath and insufficient sampling time intervals, resulting in loss of information about rain-
fall values and rainfall types. The Indonesian region is characterized by a high variability in rainfall and strong convective activity. Precipitation events outside these satellite obser-
vation windows directly resulted in monthly and seasonal statistical errors. In addition, the rainfall data used in this study have a limited sampling representation, where 0.25
◦
× 0.25
◦
3B43 data are represented by only one rain gauge. Previous studies show that convective rainfall predominantly controls tropical rainfall
peaks in Indonesia Kubota, Numaguti, and Emori 2004; Mori et al. 2004; Tabata et al. 2011; Prasetia, As-syakur, and Osawa 2013. Liao and Meneghini 2009, using ground-
based radar, found particularly in heavy rain underestimation of TRMM PR attenuation for convective rain, while stratiform rain was more accurately corrected. In addition, when
convective rain is about 50–70, this is possibly due to the overestimate of rainfall by TMI Nakazawa and Rajendran 2004. Therefore, differences in rainfall in this region may be
caused by a succession of convective and stratiform rain types throughout the year. Both convective and stratiform rain types dominate in the Indonesian archipelago Schumacher
and Houze 2003; Yulihastin and Kodama 2010; Prasetia, As-syakur, and Osawa 2013.
5. Summary and conclusions
